CN120492875B - Pump room automatic monitoring and early warning method and system - Google Patents

Abstract

The present application relates to the field of monitoring and early warning technology, and discloses a method and system for automated monitoring and early warning of a pump room. The method includes: collecting hydraulic pulsation, cavitation acoustic emission, and bearing temperature rise data through multiple sensors; obtaining a pump efficiency attenuation vector based on characteristics optimized by an environmental adaptive algorithm; predicting the health status of the equipment using a hydraulic model; using a fault propagation network for deep learning classification; and combining water inflow-pump efficiency coupling judgment to achieve a four-level graded early warning. The present application solves the technical problem in the prior art of being unable to accurately identify gradual failures and multi-parameter coupling failure modes of pump room equipment under complex underground working conditions, thereby improving the accuracy of fault prediction and the timeliness of early warning in underground drainage pump rooms.

Description

Pump room automatic monitoring and early warning method and system

Technical Field

The application relates to the technical field of monitoring and early warning, in particular to a pump room automatic monitoring and early warning method and system.

Background

The existing pump room automatic monitoring and early warning technology mainly adopts a monitoring mode based on single sensor data and fixed threshold judgment, and acquires operation parameters by arranging devices such as a water level sensor, a pressure sensor, a temperature sensor and the like in a pump room, and triggers an alarm signal when the monitoring parameters exceed a preset threshold. The traditional system generally adopts a Programmable Logic Controller (PLC) or a Distributed Control System (DCS) as a control core, and realizes the equipment state monitoring and alarming functions by using simple logic judgment and comparison operation. The monitoring and early warning method in the prior art mainly comprises the modes of equipment performance evaluation based on an empirical formula, fault identification based on statistical analysis, diagnosis decision based on an expert system and the like, and the methods realize the automatic monitoring and basic early warning functions of pump room equipment to a certain extent.

However, the traditional single sensor monitoring mode cannot fully reflect the complex running state of the equipment, especially in the underground complex environment, equipment faults often appear as multi-parameter coupling changes, and the problem of false alarm and false missing is easily caused by single-parameter threshold judgment. Secondly, the alarm mechanism with fixed threshold lacks self-adaptive capability to environmental changes, and changes of downhole temperature, humidity, water quality and other environmental factors can influence the accuracy of the sensor and the performance of equipment, but the existing system cannot dynamically adjust the judgment standard to adapt to the environmental changes. In addition, the traditional method mainly relies on post alarm rather than predictive maintenance, and can only alarm when obvious faults of equipment occur, gradual faults and potential risks of the equipment cannot be identified in advance, so that the maintenance is passive and the equipment is damaged irreversibly.

Based on the analysis, the problem in the prior art is that the multi-dimensional data fusion and intelligent analysis capability are lacked, and the failure mode of pump house equipment under the underground complex working condition cannot be accurately identified and predicted. Specifically, the prior art lacks a hydraulic-mechanical coupling monitoring means aiming at underground special environments, cannot effectively capture characteristic information of typical underground faults such as cavitation, abrasion, lubrication failure and the like, lacks an environment-adaptive variable selection and characteristic optimization algorithm, cannot dynamically adjust the weight and importance of monitoring parameters according to underground environment changes, lacks a fault mode identification capability based on deep learning, cannot learn and identify complex fault evolution rules and multiple fault coupling modes, lacks an early warning mechanism of coupling analysis of water inflow and equipment performance, and cannot comprehensively consider the influence of hydrologic condition changes on the running state of the equipment, so that early warning accuracy and timeliness are insufficient.

Disclosure of Invention

The application provides a pump room automatic monitoring and early warning method and system, which are used for solving the technical problem that gradual faults and multi-parameter coupling fault modes of pump room equipment under complex underground working conditions cannot be accurately identified in the prior art, and improving the accuracy of underground drainage pump room fault prediction and the timeliness of early warning.

The application provides a pump room automatic monitoring and early warning method, which comprises the steps of carrying out multidimensional data acquisition and processing on an underground drainage pump room through a hydraulic vibration sensor, an impeller cavitation acoustic sensor and a bearing temperature rise sensor to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data;

performing characteristic optimization processing on the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data according to a downhole environment self-adaptive variable selection algorithm to obtain a pump efficiency attenuation characteristic vector;

Performing fault mode modeling processing on the pump efficiency attenuation feature vector through a hydraulic attenuation model to obtain water pump health state prediction data;

performing deep learning training treatment on the water pump health state prediction data according to the underground water pump fault propagation network to obtain a fault identification classification result;

and carrying out grading early warning processing on the running state of the pump house through water inflow-pump efficiency coupling judgment on the fault identification classification result to obtain a four-stage early warning output signal.

In a second aspect, the present application provides a pump room automation monitoring and early warning system, the pump room automation monitoring and early warning system comprising:

the acquisition module is used for carrying out multidimensional data acquisition and processing on the underground drainage pump room through the hydraulic vibration sensor, the impeller cavitation acoustic sensor and the bearing temperature rise sensor to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data;

The optimization module is used for carrying out characteristic optimization processing on the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data according to an underground environment self-adaptive variable selection algorithm to obtain a pump efficiency attenuation characteristic vector;

the modeling module is used for carrying out fault mode modeling processing on the pump efficiency attenuation characteristic vector through a hydraulic attenuation model to obtain water pump health state prediction data;

The training module is used for performing deep learning training treatment on the water pump health state prediction data according to the underground water pump fault propagation network to obtain a fault identification classification result;

and the early warning module is used for carrying out grading early warning processing on the operation state of the pump house through water inflow and pump efficiency coupling judgment on the fault identification and classification result to obtain a four-stage early warning output signal.

In a third aspect, the pump room automation monitoring and early warning device comprises a memory and at least one processor, wherein instructions are stored in the memory, and the at least one processor calls the instructions in the memory so that the pump room automation monitoring and early warning device can execute the pump room automation monitoring and early warning method.

In a fourth aspect, a computer readable storage medium having instructions stored therein that, when executed on a computer, cause the computer to perform the pump house automation monitoring and early warning method described above is provided.

According to the technical scheme provided by the application, the hydraulic vibration sensor, the impeller cavitation acoustic sensor and the bearing temperature rise sensor are used for collecting multidimensional data of the underground drainage pump room, so that synchronous monitoring of hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data is realized, and compared with a single sensor monitoring mode in the prior art, multiple fault characteristic information of underground pump room equipment under complex hydrogeology conditions can be comprehensively captured, and the problems of information loss and misjudgment caused by single parameter monitoring are effectively avoided. The underground environment self-adaptive variable selection algorithm can dynamically adjust the parameter weight of the sensor according to the real-time change of underground temperature, humidity and pH value, overcomes the limitation that the traditional fixed weight method cannot adapt to the environment change, ensures that key fault characteristics can still be accurately extracted in the underground severe environment, and remarkably improves the environment adaptability and accuracy of the characteristic extraction. The hydraulic attenuation model establishes a pump efficiency attenuation rule based on cavitation loss coefficients and impeller abrasion loss coefficients, can accurately describe the influence mechanism of underground specific cavitation, abrasion, lubrication failure and other fault modes on equipment performance, provides a solid theoretical basis for fault prediction, and has stronger physical significance and prediction precision compared with the traditional empirical formula.

The underground water pump fault propagation network adopts a deep learning technology to train and process the water pump health state prediction data, can automatically learn and identify complex fault modes and evolution rules thereof, breaks through the technical bottleneck that the traditional rule-based expert system cannot process complex nonlinear relations, realizes accurate classification and identification of normal states, slight faults, medium faults and serious faults, and has the classification accuracy remarkably superior to that of the traditional method. The water inflow-pump efficiency coupling judgment mechanism creatively analyzes the association between the water inflow working condition and the equipment performance, can comprehensively consider the influence of underground hydrologic condition change on the operation of the equipment, realizes the transition from passive post alarm to active prediction and early warning, and provides a hierarchical management strategy for fault risks of different degrees by four-level early warning output signals, so that maintenance decisions are more scientific and reasonable. Particularly in the specific application field of underground drainage of coal mines, the algorithm model fully considers the specificity of underground environment and the complexity of water pump faults, and the accuracy of fault identification and the timeliness of early warning are remarkably improved through the organic combination of multi-sensor data fusion, environment self-adaptive processing and deep learning analysis.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of an embodiment of a pump house automation monitoring and early warning method according to an embodiment of the present application;

FIG. 2 is a schematic diagram of an embodiment of an automated pump house monitoring and early warning system according to an embodiment of the present application;

fig. 3 is a schematic block diagram of a pump house automation monitoring and early warning device in an embodiment of the invention.

Detailed Description

The embodiment of the application provides a pump room automatic monitoring and early warning method and system. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

For ease of understanding, a specific flow of an embodiment of the present application is described below, referring to fig. 1, and an embodiment of a pump house automation monitoring and early warning method in an embodiment of the present application includes:

S101, carrying out multidimensional data acquisition and processing on an underground drainage pump room through a hydraulic vibration sensor, an impeller cavitation acoustic sensor and a bearing temperature rise sensor to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data;

Step S102, carrying out characteristic optimization processing on hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data according to an underground environment self-adaptive variable selection algorithm to obtain a pump efficiency attenuation characteristic vector;

Step S103, carrying out fault mode modeling processing on the pump efficiency attenuation feature vector through a hydraulic attenuation model to obtain water pump health state prediction data;

Step S104, deep learning training processing is carried out on the water pump health state prediction data according to the underground water pump fault propagation network, and a fault recognition classification result is obtained;

And step 105, carrying out grading early warning processing on the operation state of the pump house through water inflow-pump efficiency coupling judgment on the fault identification classification result to obtain a four-stage early warning output signal.

It can be understood that the execution body of the application can be a pump room automation monitoring and early warning system, and can also be a terminal or a server, and the execution body is not limited in the specific description. The embodiment of the application is described by taking a server as an execution main body as an example.

Specifically, the multi-dimensional data acquisition is realized through a special sensor network of an underground drainage pump house, wherein a hydraulic vibration sensor is integrated with a piezoelectric vibration detection unit and a hydraulic pulsation measurement unit, mechanical vibration and hydraulic pulsation signals induced by water flow are monitored simultaneously, the piezoelectric vibration detection unit converts the mechanical vibration into an electric signal, the hydraulic pulsation measurement unit directly measures the pressure change of the water flow, the hydraulic pulsation amplitude is extracted through frequency domain analysis after the data fusion of the two data, and the amplitude reflects the strength change of the interaction of a water pump impeller and the water flow. The impeller cavitation acoustic sensor captures high-frequency acoustic emission signals generated by impeller cavitation by adopting a broadband acoustic emission technology, acoustic signals of a specific frequency band are generated when cavitation bubbles are formed on the surface of the impeller of the water pump and are broken, the acoustic signals are converted into digital signals by the sensor, and acoustic emission intensity values are calculated and directly reflect the severity of the impeller cavitation. The bearing temperature rise sensor measures the temperature change of the bearing seat through the thermistor, and when the bearing is worn or has poor lubrication, extra heat can be generated to cause the temperature rise, and the sensor continuously records temperature data and calculates the temperature rise rate.

The underground environment self-adaptive variable selection algorithm firstly establishes a water-inrush working condition classification matrix, divides the underground working condition into different grades according to real-time water inflow, and each grade corresponds to a specific water pump operation parameter range. The algorithm extracts key inflection point parameters on a pump lift-flow characteristic curve, including an optimal efficiency point flow, a shut-off lift, a maximum flow and an efficiency drop inflection point, which describe the performance characteristics of the pump under different working conditions. The principal component analysis algorithm calculates the correlation between each sensor parameter and the pump efficiency attenuation to generate a weight matrix, and each weight coefficient represents the influence degree of the corresponding sensor parameter on the pump efficiency change. Because temperature, humidity and water quality changes in the underground environment can influence the sensor precision, the algorithm dynamically corrects the weight matrix, and the correction process considers the influence of temperature changes on the sensitivity of the piezoelectric sensor, the influence of humidity changes on the performance of electronic elements and the influence of pH value changes on the response characteristic of the acoustic sensor. The corrected weight coefficient is used for screening the most representative characteristic variable to form a pump efficiency attenuation characteristic vector.

The hydraulic attenuation model establishes a pump efficiency attenuation law based on a water pump similarity law and an impeller abrasion theory, and the model decomposes pump efficiency attenuation into two main components, namely cavitation loss and abrasion loss. The cavitation loss coefficient is calculated by the cavitation coefficient, the cavitation coefficient reflects the relationship between the inlet condition of the water pump and the geometric parameters of the impeller, and when the cavitation coefficient is reduced, the cavitation phenomenon is aggravated. The abrasion loss coefficient is calculated based on the corrosiveness of water and the concentration of particles, and the higher the concentration of the particles in water is, the more serious the impeller scour and abrasion is, and the metal corrosion can be accelerated when the pH value of the water deviates from neutrality. The polynomial fitting is used for describing the degradation process of the pump lift-flow characteristic curve along with time, and the fitting coefficient reflects the change trend of the shape of the characteristic curve. The bearing water lubrication theory analyzes the lubrication state change of the bearing in the underground high-humidity environment, and when moisture enters the bearing cavity, lubricating grease can be diluted, the lubrication effect is reduced, and the temperature of the bearing is increased and vibration is aggravated.

The underground water pump fault propagation network adopts a multi-layer neural network architecture to process complex fault mode identification tasks, and an input layer receives multidimensional data from pump efficiency attenuation characteristic vectors, wherein the multidimensional data comprises key parameters such as impeller cavitation acoustic emission intensity, bearing temperature rise rate, hydraulic vibration main frequency and the like. The hidden layer uses a ReLU activation function to perform nonlinear transformation, the ReLU function sets the negative input to zero and keeps the positive value unchanged, so that the gradient vanishing problem is effectively avoided and the network convergence is accelerated. The network adjusts the connection weight through a back propagation algorithm, the cross entropy loss function is used for error calculation of classification tasks, the mean square error loss function is used for error calculation of regression tasks, and the two functions are weighted and combined to form a total loss function to guide network training. The four nodes of the output layer respectively correspond to a normal state, a slight fault, a medium fault and a serious fault, and each node outputs a probability value of the corresponding state.

The water inflow-pump efficiency coupling judgment establishes a multidimensional judgment matrix, wherein matrix rows represent different water inflow grades, columns represent different pump efficiency ranges, and matrix element values represent early warning grades under corresponding working conditions. When the water inflow is at a low level and the pump efficiency is higher than a set threshold, the water pump is judged to be in primary early warning, and the water pump operates in an ideal working condition and has good equipment state. When the water inflow is increased or the pump efficiency is reduced to a medium range, a secondary early warning is triggered, which indicates that the equipment performance begins to decay and needs to be concerned. When the water inflow reaches a higher level and the pump efficiency is obviously reduced, three-level early warning is triggered, and at the moment, the equipment has obvious fault risk and needs to be overhauled in time. Triggering a four-level early warning when the water inflow exceeds a design limit or the pump efficiency falls below a dangerous threshold, and immediately taking emergency measures when equipment faces serious fault threat. The influence degree of the water inflow change on the pump efficiency is considered in the coupling weight calculation, and when the water inflow is increased sharply, the early warning level needs to be improved even if the pump efficiency is not reduced obviously temporarily, because the equipment load is increased under the working condition of large water inflow, and the fault risk is increased.

In a specific embodiment, the process of executing step S101 may specifically include the following steps:

the method comprises the steps of performing signal acquisition processing on a piezoelectric vibration detection unit and a water pressure pulsation measurement unit at a water inlet of a water pump based on a hydraulic vibration sensor to obtain hydraulic excitation frequency spectrum data in a frequency band of 0.1Hz-5 kHz;

carrying out amplitude extraction processing on the hydraulic excitation frequency spectrum data to obtain hydraulic pulsation amplitude;

Carrying out broadband acoustic emission detection treatment on a 20kHz-200kHz frequency band of the impeller shell according to the impeller cavitation acoustic sensor to obtain a bubble collapse acoustic characteristic signal;

performing acoustic emission intensity calculation processing on the bubble collapse acoustic characteristic signals to obtain cavitation acoustic emission intensity;

And carrying out 0.1 ℃ resolution monitoring treatment on the temperature change of the bearing based on the bearing temperature rise sensor to obtain bearing temperature rise data.

Specifically, the medium hydraulic vibration sensor is deployed at a key position of a water inlet of the water pump, and two independent measuring units are integrated inside to realize a collaborative monitoring function. The piezoelectric vibration detection unit is made of piezoelectric ceramic materials, when the water pump operates to generate mechanical vibration, the piezoelectric materials are subjected to stress to deform, the internal charge distribution of the materials changes in the deformation process to generate corresponding voltage signals, and the amplitude of the voltage signals is in a proportional relation with the vibration intensity. The water pressure pulsation measuring unit adopts a strain type pressure sensor, an elastic diaphragm in the sensor generates tiny displacement under the action of water flow pressure change, and the displacement is converted into resistance change through a strain gauge to finally output a voltage signal. The two units synchronously acquire data and then carry out amplification and filtering processing through a signal conditioning circuit, and the sampling frequency is set to be ten thousand times per second so as to ensure the capture of the complete signal characteristics in the frequency range from zero hertz to five kilohertz. The digital signal processor performs fast Fourier transform on the acquired time domain signals, converts the time domain vibration signals into frequency domain frequency spectrum data, and the frequency spectrum data comprises amplitude and phase information of each frequency component, wherein the amplitude information reflects vibration energy distribution conditions under corresponding frequencies. The amplitude extraction process of the hydraulic excitation frequency spectrum data adopts a method combining a peak detection algorithm and power spectrum density analysis. The peak detection algorithm identifies local maximum points in the frequency spectrum data, wherein the peak points correspond to main excitation frequency components in the operation process of the water pump, and the main excitation frequency components comprise impeller passing frequency, shaft frequency and harmonic frequencies of each order. And (3) analyzing and calculating the energy density distribution in each frequency band by power spectrum density analysis, and obtaining the total energy value in the specific frequency band through integral operation. The hydraulic pulsation amplitude is obtained by carrying out weighted average calculation on the amplitude corresponding to the main excitation frequency, the weight coefficient is determined according to the influence degree of each frequency component on the hydraulic performance of the water pump, the weight of the impeller passing frequency is highest because the impeller passing frequency directly reflects the interaction strength of the impeller and the water flow, and the weight of the shaft frequency and the harmonic wave thereof is relatively low but the contribution of the shaft frequency and the harmonic wave to the whole vibration level still needs to be considered.

The impeller cavitation acoustic sensor is arranged on the surface of the impeller shell, and an acoustic signal generated by cavitation phenomenon is monitored by adopting a broadband acoustic emission technology. When cavitation occurs, dissolved gas in water is separated out in a low-pressure area to form bubbles, the bubbles instantaneously burst to release a large amount of energy when moving to a high-pressure area along with water flow, and the burst process generates high-frequency sound waves with the frequency range of twenty-hundred thousand hertz. The acoustic sensor adopts a piezoelectric ceramic transducer to convert sound wave vibration into an electric signal, and the surface of the transducer is plated with a special coating to enhance the sensitivity to high-frequency signals. The signal conditioning circuit comprises a pre-amplifier, a band-pass filter and an automatic gain control module, wherein the pre-amplifier amplifies weak acoustic emission signals to a processable range, the band-pass filter filters interference noise outside a frequency band, and the automatic gain control module dynamically adjusts amplification factors according to signal intensity to prevent signal saturation. The characteristic parameters of bubble collapse, including the occurrence frequency, duration and energy release intensity of the collapse event, are extracted from the digitally processed acoustic emission signals through a time-frequency analysis algorithm.

The acoustic emission intensity of the bubble collapse acoustic characteristic signal is calculated by adopting a mode of combining an energy integration method and an event counting method. The total energy of the signal in the time window is calculated by an energy integration method, and the total energy is obtained by carrying out time integration on the square of the amplitude of the acoustic emission signal, and the integral result reflects the total energy released by the collapse of cavitation bubbles. The event counting method counts the number of acoustic emission events exceeding a preset threshold, each bubble burst corresponds to one acoustic emission event, and the number of the events and the cavitation intensity are in positive correlation. The acoustic emission intensity comprehensively considers two factors of energy density and event frequency, and a final cavitation acoustic emission intensity value is obtained through weighted calculation, the value quantitatively describes the severity of cavitation phenomenon on the surface of the impeller, and the greater the value, the stronger the destructive effect of cavitation on the impeller is indicated.

The bearing temperature rise sensor adopts a platinum resistance thermometer to realize high-precision temperature measurement, the resistance value of the platinum resistance material is in linear relation with the temperature, and the resistance change amount corresponding to each temperature change of one degree celsius is fixed and has good repeatability. The sensor is packaged in the waterproof explosion-proof shell and is directly contacted with the outer surface of the bearing seat, and the temperature change in the bearing is sensed in a heat conduction mode. The measuring circuit adopts a four-wire system connection method to eliminate the influence of lead resistance on measurement precision, the constant current source provides stable excitation current, and the voltage measuring circuit detects voltage changes at two ends of the platinum resistor and converts the voltage changes into digital temperature values. The temperature data acquisition period is set to be ten times per second, and the influence of short-term temperature fluctuation is eliminated through a moving average filtering algorithm in the continuous monitoring process. The temperature rise data of the bearing is obtained through the calculation of the difference value between the current temperature value and the reference temperature value, the reference temperature value is the steady-state temperature of the bearing in the normal running state, and the temperature rise data is updated in real time and stored in the circulating buffer zone.

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

Performing low-flow, medium-high-flow and high-flow grade classification treatment on underground water inflow working conditions based on the water inflow grading matrix to obtain water inflow working condition classification labels;

carrying out key inflection point parameter extraction processing on a pump lift-flow characteristic curve according to the water inflow working condition classification label to obtain the inflection point parameters of optimal efficiency point flow, turn-off lift, maximum flow and efficiency reduction;

inputting the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data into a principal component analysis algorithm to perform correlation weight calculation processing, so as to obtain a sensor parameter weight matrix;

Carrying out dynamic correction processing on the underground temperature, humidity and pH value change on the sensor parameter weight matrix to obtain an environment self-adaptive weight coefficient;

And screening and optimizing the core characteristic variable based on the environment self-adaptive weight coefficient to obtain the pump efficiency attenuation characteristic vector.

Specifically, the water inflow grading matrix establishment process firstly determines grading standards based on underground hydrogeological conditions and water pump design parameters, and the matrix adopts four rows and four columns of structures to respectively correspond to four grades of low flow, medium and high flow. The low flow rate level corresponds to the normal water seepage condition in the well, the water inflow is usually in the range of zero to fifty cubic meters per hour, and the water pump is relatively low in efficiency under the partial load state and light in equipment burden. The medium-low flow level corresponds to a slight water inflow working condition, the water inflow is in the range of fifty to one hundred cubic meters per hour, and the efficiency of the water pump starts to enter a more ideal operation interval to be gradually improved. The medium-high flow level corresponds to a medium water inflow working condition, the water inflow is in the range of one hundred to two hundred cubic meters per hour, and the efficiency of the water pump reaches a higher level near the design working condition but the equipment load is obviously increased. The high flow level corresponds to a large or sudden water gushing condition, the water gushing amount exceeds two hundred cubic meters per hour, the efficiency of the water pump is reduced in an overload state, and the equipment faces a large fault risk. The grading processing is realized by comparing the real-time water inflow data with a preset threshold value, when the water inflow exceeds the upper limit threshold value of a certain grade, the water inflow is automatically switched to a higher grade, and each grade corresponds to a specific water inflow working condition classification label for subsequent parameter extraction and analysis processing.

The pump lift-flow characteristic curve describes the lift output capability of the pump under different flows, and the curve shape reflects the hydraulic performance characteristics of the pump. The key inflection point parameter extraction process firstly selects corresponding characteristic curve data according to the water gushing working condition classification labels, and the difference of the running characteristics of the water pump under different water gushing working conditions needs to be treated respectively. The optimal efficiency point flow is determined through the peak point of the efficiency curve, and the point corresponds to the flow value when the efficiency of the water pump is highest, so that the optimal efficiency point flow is an ideal working point of the water pump design. The off-lift refers to the maximum lift output of the water pump in a zero flow state, and is obtained through the intercept of a characteristic curve on a vertical axis, wherein the parameter reflects the maximum pressure head capacity of the water pump. The maximum flow refers to the maximum flow output of the water pump in the zero-lift state, and is obtained through the intercept of the characteristic curve on the transverse axis, and the parameter reflects the maximum drainage capacity of the water pump. The efficiency-decreasing inflection point is the inflection point at which the efficiency curve decreases significantly from the peak value, and is determined by curve slope change analysis, and marks the beginning of the deviation of the water pump from the optimal working area. The inflection point parameter extraction adopts a method of combining numerical differentiation and curve fitting, firstly, a characteristic curve is subjected to cubic spline fitting to obtain a smooth continuous function, and then the accurate position of each inflection point is determined through calculation of a first derivative and a second derivative.

When the principal component analysis algorithm processes multidimensional sensor data, firstly, data standardization processing is carried out to eliminate the influence of different physical dimensions, and the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data are converted into standardized data with zero mean and one variance. Covariance matrix calculation describes a linear correlation between sensor parameters, matrix element values reflect the degree of correlation of two parameters, positive values represent positive correlation and negative values represent negative correlation. The eigenvalue decomposition decomposes the covariance matrix into eigenvalues and eigenvectors, the eigenvalue magnitude reflects the variance contribution rate of the corresponding principal component, and the eigenvector reflects the weighting coefficient of each sensor parameter on the principal component. The weight matrix is composed of eigenvectors arranged according to the eigenvalue size, and the first few principal components usually contain the principal information of the data, and the subsequent components are mainly noise. The correlation weight calculation is determined by the correlation coefficient of each sensor parameter and pump efficiency attenuation, the larger the absolute value of the correlation coefficient is, the more obvious the influence of the parameter on the pump efficiency change is, and the higher the value is correspondingly set for the weight coefficient.

The dynamic correction of the sensor parameter weight matrix considers the influence of downhole environmental factors on the sensor performance, and the correction process is performed based on real-time monitoring data of environmental parameters. The temperature change in the well mainly affects the sensitivity of the piezoelectric sensor and the working characteristics of electronic devices, the piezoelectric constant of the piezoelectric material is reduced when the temperature is increased, so that the amplitude of the output signal of the sensor is reduced, and the correction coefficient is calculated by the temperature coefficient and the difference value between the current temperature and the standard temperature. The change of the underground humidity influences the insulating performance and the signal transmission quality of the electronic element, leakage current and signal interference are easy to generate in a high humidity environment, and the correction coefficient is determined according to the relative humidity value and the humidity sensitivity parameter. The pH value change of water quality affects the response characteristic of an acoustic sensor and the corrosion degree of metal parts, the acidic or alkaline environment can change the surface characteristic of the sensor and affect the sound wave propagation effect, and the correction coefficient is obtained through the calculation of the degree of pH value deviation from a neutral value and the corrosion constant of materials. The dynamic correction process multiplies the original weight coefficient and each environment correction coefficient to obtain a corrected weight value, and the correction process is performed in real time to ensure that the weight matrix can accurately reflect the reliability of the sensor under the current environment condition.

The core feature variable screening is optimized based on the environment self-adaptive weight coefficient, and the screening process aims at extracting the feature parameters with the most representation and reliability from the multidimensional sensor data. The weight threshold value is set for judging the importance degree of the feature variable, the variable with the weight coefficient higher than the threshold value is selected as the core feature, and the variable with the weight coefficient lower than the threshold value is regarded as redundant information or noise interference. The characteristic variables comprise key parameters such as impeller cavitation acoustic emission intensity, bearing temperature rise rate, hydraulic vibration main frequency offset, pump efficiency real-time value, lift deviation, flow fluctuation coefficient and the like, and each parameter reflects the running state and health degree of the water pump from different angles. The screening optimization adopts a stepwise regression method, firstly, a feature variable with the highest weight is selected as a basis, then other feature variables are gradually added, the improvement degree of the prediction precision is evaluated, and when the new variable cannot significantly improve the prediction effect, the addition is stopped. The pump efficiency attenuation characteristic vector is formed by arranging screened core characteristic variables according to the weight, and the length of the vector is determined according to the prediction accuracy requirement and the calculation complexity balance, and generally comprises eight to twelve characteristic variables.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

based on cavitation acoustic emission intensity in the pump efficiency attenuation characteristic vector and impeller cavitation acoustic sensor data, performing cavitation coefficient calculation processing to obtain cavitation loss coefficients;

carrying out abrasion loss coefficient calculation according to the water quality pH value and the particle concentration parameter in the pump efficiency attenuation characteristic vector to obtain an impeller abrasion loss coefficient;

inputting cavitation loss coefficients and impeller abrasion loss coefficients into a water pump similarity law to perform modeling processing on a pump efficiency attenuation law, so as to obtain pump efficiency time-varying attenuation data;

Performing polynomial fitting calculation on the pump efficiency time-varying attenuation data to obtain a pump lift-flow characteristic curve degradation parameter;

And carrying out comprehensive evaluation treatment on the health state based on the degradation parameters of the pump lift-flow characteristic curve and the bearing water lubrication theory to obtain the predicted data of the health state of the water pump.

Specifically, the cavitation coefficient calculation is based on the hydrodynamic condition at the impeller inlet and cavitation acoustic emission data for quantitative analysis, and the cavitation coefficient reflects an important index of the cavitation resistance of the water pump. The method comprises the steps of firstly determining cavitation occurrence intensity and cavitation occurrence frequency by using real-time data acquired by an impeller cavitation acoustic sensor, and comprehensively evaluating cavitation acoustic emission intensity through energy integration and occurrence frequency of acoustic emission events. The cavitation coefficient is defined as the ratio of the effective cavitation margin at the impeller inlet to the dynamic head, the effective cavitation margin being determined by inlet pressure, steam pressure and velocity head calculations, and cavitation beginning to occur when the effective cavitation margin falls below a threshold value. The cavitation loss coefficient is established by the ratio relation of the cavitation coefficient and the critical cavitation coefficient, and when the actual cavitation coefficient is lower than the critical value, the cavitation loss coefficient is increased sharply, and the loss coefficient corresponding to the serious cavitation phenomenon is larger. The calculation process also needs to consider the influence of water quality conditions on the cavitation characteristics, and the cavitation characteristics of water can be changed by the content of dissolved gas, the water temperature and the impurity concentration, and the factors are included in the calculation process in the form of correction coefficients.

The abrasion loss coefficient calculation is quantitatively evaluated based on water quality analysis data and particle abrasion theory, and underground water quality is complex and changeable to cause abrasion damage to the impeller of the water pump to different degrees. The pH value of water reflects the acid-base degree of water, the electrochemical corrosion process of the metal material can be accelerated by an acidic environment, the passivation film on the metal surface can be damaged by an alkaline environment, and the corrosion effect on the impeller material is stronger as the pH value deviates from neutrality. Particle concentration parameters are obtained by measurement of a nephelometer and a particle counter, and include the number density, the particle size distribution and the hardness characteristics of particles, and the scouring abrasion effect of hard particles on the surface of an impeller is most remarkable. The abrasion loss coefficient is calculated by adopting a correction form of an Alchard abrasion law, and the comprehensive influence of a plurality of factors such as particle hardness, impact speed, contact time, material performance and the like is considered. The synergistic effect of corrosive wear is described by the coupled relationship of the corrosion and wear factors, which when both pH and particle concentration are high, results in more severe material loss than would be the case alone. The calculation process also needs to consider the accumulated effect of the running time, the abrasion loss has a nonlinear growth trend along with the running time, and the initial growth is slow and the later growth is accelerated to be worsened.

The water pump similarity law describes the performance relation of the geometrically similar water pump under different working conditions, and provides a theoretical basis for modeling the pump efficiency attenuation law. The law of similarity includes three basic relationships of flow similarity, lift similarity and power similarity, and the relationships deviate under the influence of cavitation and abrasion to be corrected. The pump efficiency attenuation law modeling introduces cavitation loss coefficient and abrasion loss coefficient as correction factors into a similar law, and the corrected efficiency relation can describe the performance attenuation process of the water pump under the actual running condition. Cavitation loss mainly affects the flow coefficient and the lift coefficient of the water pump, and cavitation causes the effective flow area of the impeller to be reduced and the pressure pulsation to be increased, so that the hydraulic efficiency of the water pump is reduced. Wear loss primarily affects impeller geometry and surface roughness, and wear results in reduced blade thickness, changes in blade angle, and increased surface roughness, which can reduce the hydraulic performance and volumetric efficiency of the impeller. The modeling process adopts a time sequence analysis method, the efficiency attenuation is expressed as a function of time, and the function parameters are determined through regression analysis of historical operation data. The pump efficiency time-varying attenuation data comprise efficiency predicted values and confidence intervals at different moments, and a quantitative basis is provided for equipment maintenance decision.

The polynomial fitting is used for describing the degradation process of the pump lift-flow characteristic curve along with time, and the fitting process is performed based on pump efficiency time-varying attenuation data and a hydraulic theory. Characteristic degradation is mainly manifested by overall downward curve movement, slope changes and shape distortions, which reflect the decay process of impeller geometry and hydraulic performance. Fitting adopts a piecewise polynomial method, a characteristic curve is divided into a low flow section, a medium flow section and a high flow section to be respectively processed, and each section adopts a cubic polynomial to describe the relation between the lift and the flow. The change rule of the polynomial coefficient along with time is determined through time sequence analysis, the constant term reflects the integral position change of the curve, the first-order term coefficient reflects the slope change of the curve, and the higher-order term coefficient reflects the bending degree change of the curve. The fitting quality is evaluated by determining coefficients reflecting the interpretation degree of the fitting curve on the actual data and residual analysis for verifying the distribution characteristics and systematic deviation of the fitting error. The degradation parameters comprise key indexes such as maximum lift attenuation rate, optimal efficiency point offset, high-efficiency interval shrinkage degree, curve distortion index and the like, and the parameters quantitatively describe the degradation degree and the development trend of the characteristic curve.

The bearing water lubrication theory analyzes the lubrication state change and the performance attenuation law of the bearing in the underground high-humidity environment, and provides theoretical support for health state assessment. Moisture in the underground environment easily enters the bearing cavity to dilute lubricating grease and change the lubricating property, and the friction coefficient, the bearing capacity and the service life of the bearing can be obviously changed under the condition of water lubrication. The lubrication state evaluation is comprehensively judged through bearing temperature rise data, vibration characteristics and lubricating grease analysis results, the bearing temperature rise is stable and the vibration level is lower in the normal lubrication state, and the temperature rise is accelerated and the vibration is enhanced in the poor lubrication state. The water lubrication theory considers the influence of water film thickness, viscosity change and boundary lubrication effect on bearing performance, and metal direct contact and abrasion are aggravated when the water film thickness is insufficient. And (3) comprehensively evaluating the health state, wherein the degradation parameter of the lift-flow characteristic curve and the lubrication state parameter of the bearing are subjected to weighted fusion, and the evaluation weight is determined according to the influence degree of each parameter on the overall performance of the equipment. The evaluation result is expressed in the form of a health index, the health range is from zero to one hundred, the higher the value is, the better the equipment state is, and the corresponding maintenance suggestion is triggered when the health is lower than a set threshold value. The water pump health state prediction data comprises information such as current health degree, future week health degree prediction value, main attenuation factor identification and maintenance advice.

In a specific embodiment, the process of executing step S104 may specifically include the following steps:

Inputting the predicted data of the water pump health state into an input layer of a multi-layer neural network for data receiving processing to obtain input data corresponding to the cavitation acoustic emission intensity of the impeller, the temperature rise rate of the bearing and the main frequency of hydraulic vibration;

performing ReLU activation function calculation processing on hidden layer neuron nodes based on input data to obtain a nonlinear feature mapping result;

performing fault state classification processing on the four nodes of the output layer according to the nonlinear feature mapping result to obtain classification probabilities of normal states, slight faults, medium faults and serious faults;

The classified probability is input into a cross entropy loss function and a mean square error loss function to carry out weighted combination loss calculation processing, and a network training loss value is obtained;

And carrying out back propagation optimization processing on the network parameters based on the network training loss value to obtain a fault identification classification result.

Specifically, the input layer of the multi-layer neural network is designed to receive and preprocess the water pump health state prediction data, and the input layer comprises twelve neuron nodes corresponding to different characteristic parameters respectively. Firstly, the water pump health state prediction data is subjected to data standardization processing, parameters of different physical dimensions and numerical ranges are converted into standardized numerical values with zero standard deviation as one mean value, and the fact that all characteristic parameters have the same weight basis in network training is ensured. The impeller cavitation acoustic emission intensity data reflects the severity of impeller surface cavitation phenomenon, the acoustic emission intensity value is compared with a historical reference value in the input processing process, the relative change rate is calculated, and the change rate value is subjected to logarithmic transformation to reduce the dynamic range of the data so as to facilitate network processing. The temperature rise rate data of the bearing is obtained by carrying out numerical differential calculation on the bearing temperature time sequence, the differential process adopts a central differential format to improve the calculation precision, the temperature rise rate reflects the change trend of the health state of the bearing, and positive values represent temperature rise negative values represent temperature drop. The hydraulic vibration main frequency data is extracted from the vibration time domain signal through fast Fourier transformation, the main frequency value reflects the main excitation characteristic of the operation of the water pump, and the main frequency offset is determined through the calculation of the difference value between the main frequency offset and the designed main frequency. The input layer carries out linear weighted combination on each characteristic parameter, and the weight coefficient is automatically adjusted and optimized in the network training process.

ReLU activation function calculation of hidden layer neuron nodes is a key step of nonlinear transformation of a network, and ReLU functions are defined as that the output is equal to the input value when the input value is larger than zero and the output is zero when the input value is smaller than or equal to zero. The activation function calculation process firstly carries out linear transformation on input data, each hidden layer neuron receives a weighted signal from an input layer, the weighted sum is realized through matrix multiplication operation, and element values of a weight matrix are continuously updated in the training process. The linear transformation result is added with the bias term and then is input into the ReLU activation function for nonlinear processing, and the ReLU function has the advantages of being simple in calculation, capable of effectively avoiding gradient disappearance, and beneficial to breaking the symmetry of the network and improving learning ability due to the asymmetry of the function. The first hidden layer contains sixty four neuron nodes, the second hidden layer contains thirty two neuron nodes, the third hidden layer contains sixteen neuron nodes, and the design of the number of layers and the number of nodes balances the expression capacity and the computational complexity of the network. The nonlinear feature mapping result is a vector set of hidden layer neuron activation values that encode a high-level feature representation of the input data, providing an information basis for classification decisions at the output layer.

The four nodes of the output layer respectively correspond to the four health states of the water pump, and the node design is based on the actual requirements of fault diagnosis and the grading requirements of maintenance decisions. The normal state node indicates that various indexes of the water pump running in the design parameter range are normal, the slight fault node indicates that the initial fault symptom is generated but the normal running is not influenced, the medium fault node indicates that the fault has shown that the maintenance plan needs to be arranged, and the serious fault node indicates that the fault is serious and the maintenance needs to be stopped immediately. The fault state classification process converts the output values of the hidden layer into probability distributions using a Softmax activation function that ensures that the sum of the probability values of the four output nodes is equal to one and each probability value is between zero and one. The probability calculation process first performs an exponential transformation on the input value of each node, and then divides the index value of each node by the sum of the index values of all nodes to obtain a normalized probability. The classification probability reflects the confidence of the network in various fault states, with higher probability values indicating a greater likelihood that the network considers the corresponding state. The classification decision is determined by probability value comparison, and the state with the highest probability value is usually selected as the classification result of the network.

The cross entropy loss function is used for measuring the difference between the classified prediction result and the real label, and the smaller the function value is, the closer the prediction result is to the real situation. The cross entropy calculation process converts the real label into a single-hot coding form, the label of each sample is expressed as a four-dimensional vector, and the corresponding position of the correct category is zero and the rest positions are zero. The loss value is obtained by carrying out dot product operation on the real tag vector and the predictive probability vector and taking the negative logarithm, when the predictive probability is close to the real tag, the loss value is smaller, and when the predictive probability is wrong, the loss value is larger. The mean square error loss function is used for measuring the numerical precision of network output, and the calculation process carries out square operation on the difference between the prediction probability and the real label and then calculates the average value. The weighted combination loss linearly combines the cross entropy loss and the mean square error loss according to a set weight proportion, the weight proportion is determined according to the importance of classification precision and numerical precision, and the weight of the cross entropy loss is set as zero point seven mean square error loss and is set as zero point three. The combined loss function comprehensively considers the requirements of classification accuracy and prediction stability, and the network training loss value is calculated through the average loss of all training samples.

The back propagation optimization carries out iterative updating on network parameters based on a gradient descent algorithm, and the optimization target is to minimize a network training loss value. The gradient calculation is propagated from the output layer to the input layer by using a chain rule, and the gradient of each layer is determined by the product of an error signal of the current layer and the output of the previous layer. The weight updating formula subtracts the product of the learning rate and the gradient from the current weight value, the learning rate controls the step size of parameter updating, and an excessive learning rate may lead to unstable training and an excessively small learning rate may prolong the training time. The optimization process adopts an Adam optimizer to combine a momentum mechanism and self-adaptive learning rate adjustment, and the Adam optimizer realizes self-adaptive parameter update by maintaining first moment and second moment estimation of the gradient.

The training process sets a maximum number of iterations of 500 rounds, each round of training comprising 4 steps of forward propagation, loss calculation, back propagation and parameter updating. Early shutdown monitors the loss variation on the validation set, and early termination of training prevents overfitting when the validation loss is not reduced for 10 consecutive rounds.

The maximum iteration number is determined by adopting a progressive test method, 100 rounds of observation loss descending trend are operated, the number of rounds when the loss value is stable is recorded, then training effects of 200 rounds, 300 rounds, 400 rounds and 500 rounds are respectively tested, and the accuracy rate of a comparison verification set and the stability of a model are compared. Through drawing a relation curve of the training round number and the verification accuracy, the verification accuracy is found to be stable around 350 rounds, and the maximum iteration times are finally set to be 500 rounds in consideration of the differences of different data batches.

The number of continuous rounds of early-stop mechanism is determined by a group comparison experiment, the same training data are trained under the early-stop settings of 5 rounds, 8 rounds, 10 rounds, 12 rounds and 15 rounds respectively, and the final training round number, the verification accuracy and the model generalization capability of each setting are recorded. The experimental results show that the 5 rounds of setting result in insufficient training and lower accuracy, the 15 rounds of setting are easy to overfit, the performance on the test set is reduced, and the 10 rounds of setting achieve the best balance between the training sufficiency and the overfit prevention.

Setting up a parameter adjustment experiment ledger, and recording key information such as training date, parameter configuration, data set version, hardware environment, training time, final accuracy, verification loss, fitting occurrence or non-fitting occurrence and the like in each parameter adjustment. The learning rate tuning adopts a gradual narrowing method, firstly 10 geometric series points are selected between 0.001 and 0.1 for coarse tuning, fine tuning is carried out in an optimal section after the optimal section is determined, and the convergence speed and the final performance are recorded after each tuning. The selection of the batch size needs to consider hardware memory limitations and training stability, and through testing of the 4 batch sizes of 16, 32, 64 and 128, the memory occupation, single-round training time, gradient update stability and final model performance under each setting are recorded. The network structure optimization adopts a control variable method, the number of hidden layers is firstly fixed to adjust the number of neurons of each layer, then the number of the neurons is fixed to adjust the number of layers, and the total network parameter, the training time and the performance index are completely recorded in each structural change.

And (3) in the training process, 1 model state and training log are stored every 10 rounds, wherein the model state and training log comprise key indexes such as current round number, training loss, verification loss, learning rate change, gradient norm and the like. When the training loss is found to continuously decrease but the verification loss starts to rise, the over-fitting detection program is started immediately, the training verification loss difference value of the last 10 rounds is compared, and if the difference value continuously expands, training is terminated in advance. If gradient explosion or disappearance occurs in the training process, the training process is processed by adjusting the learning rate or modifying the network initialization method, and each adjustment is performed by recording the adjustment reason, the adjustment method and the adjustment effect in detail.

The original data set is randomly divided into a training set, a verification set and a test set according to the proportion of 7 to 2 to 1, so that the distribution of each fault class in 3 subsets is ensured to be consistent. And (3) carrying out standardized processing on all the features before training, calculating the mean value and standard deviation of the training set, and transforming the training set, the verification set and the test set by using the same standardized parameters. The learning rate starts from 0.1 and automatically decreases to half of the original whenever the validation loss does not improve within 5 rounds until the learning rate stops adjusting to less than 0.0001. And (3) running a complete evaluation flow on the verification set after each parameter adjustment, calculating the accuracy, the precision, the recall and the F1 score, comparing the indexes with the previous optimal result, and adopting the new configuration only when the new configuration is not inferior to the optimal historical configuration on the main indexes and at least 1 index is obviously improved.

The fault identification classification result is obtained by predicting test data through a trained network, and the result comprises classification labels and corresponding confidence scores of each test sample.

In a specific embodiment, the process of executing step S105 may specifically include the following steps:

Performing coupling judgment matrix construction processing based on the fault identification classification result and the real-time pump efficiency data to obtain a water inflow-pump efficiency coupling judgment matrix;

performing primary early warning judgment processing on the condition that the pump efficiency is higher than a set threshold value and the fault is identified as a normal state according to the water inflow-pump efficiency coupling judgment matrix to obtain a green normal operation early warning signal;

Inputting the condition that the pump efficiency is in a medium range or the slight fault characteristic is detected in the water inflow-pump efficiency coupling judgment matrix into a second-level early warning judgment for processing to obtain a yellow performance reduction early warning signal;

performing three-level early warning judgment processing on the condition that the pump efficiency is in a lower range or medium fault characteristics are detected in the water inflow-pump efficiency coupling judgment matrix to obtain an orange fault risk early warning signal;

and carrying out four-level early warning judgment processing based on the condition that the pump efficiency in the water inflow-pump efficiency coupling judgment matrix is lower than a safety threshold or serious fault characteristics are detected, and obtaining a red serious fault early warning signal as a four-level early warning output signal.

Specifically, a water inflow-pump efficiency coupling judgment matrix is constructed, a comprehensive judgment framework is built based on multidimensional data fusion and decision theory, and the matrix is designed to be in a four-row four-column structure and respectively corresponds to different water inflow grades and pump efficiency range combinations. The fault identification classification result provides a qualitative judgment of the health state of the equipment, and comprises four grades of normal state, slight fault, medium fault and serious fault, and each grade corresponds to different fault probability distribution and maintenance requirements. The real-time pump efficiency data is obtained through synchronous measurement and calculation of the flowmeter, the pressure transmitter and the power meter, and the calculation process adopts an instantaneous power method to comprehensively analyze the input power, the transmission efficiency and the hydraulic power of the motor, so that an accurate numerical value reflecting the current running efficiency of the water pump is obtained. And the coupling judgment processing carries out weighted fusion on the probability value and the pump efficiency value of the fault identification result, the fusion weight is determined according to the reliability and the importance of the two types of information, and the fault identification weight is set as zero-point six pump efficiency data weight and is set as zero-point four. The matrix element values are determined by a method combining fuzzy logic and an expert system, and each element represents the comprehensive risk level under the combination of the corresponding water inflow and pump efficiency. The matrix construction process also considers the influence of the water inflow change trend on the judgment result, and the risk level needs to be improved even if the current pump efficiency is normal when the water inflow rises rapidly, so that the risk assessment can be properly reduced when the water inflow is stable or falls.

The primary early warning judging process confirms normal operation aiming at the condition that the running state of the equipment is good, and judging conditions comprise that the pump efficiency is higher than a set threshold value and that the fault is identified as the normal state. The pump efficiency threshold setting is determined based on pump design parameters and operating experience, typically set to between eighty-five and ninety percent of design efficiency, and when the measured pump efficiency exceeds the threshold, it indicates that the pump is operating well within the high efficiency interval. The normal state probability of the neural network output is required to be higher than the zero point eight and the probability of other fault states is lower than the zero point one by the fault identification normal state, and the probability distribution shows that the network has high confidence in judging the normal state. The primary early warning judgment also needs to verify that the water inflow is in the design range and the change trend is stable, and the water inflow cannot be judged to be in a normal state even if other conditions are met when the water inflow exceeds the design range or changes rapidly. The green normal operation early warning signal is presented in the form of a green indicator lamp or green characters through a display interface of the early warning system, and a time stamp and a key parameter value of a normal operation state are recorded in the monitoring log. The signal output also contains auxiliary information such as equipment operation efficiency, expected maintenance time, operation advice and the like, and provides equipment management references for operators.

The secondary early warning judgment carries out early warning on the condition that the performance of the equipment starts to decline but does not form serious threat, and the judgment conditions comprise that the pump efficiency is in a medium range or a slight fault characteristic is detected. The mid-pump efficiency range is defined as between seventy to eighty five percent of design efficiency, which indicates that the pump is still functioning properly but the efficiency is somewhat reduced and requires attention. The slight fault feature is identified by the output of the neural network that the slight fault probability is higher than the zero point five, and meanwhile, the sum of the moderate fault probability and the serious fault probability is required to be lower than the zero point three, so that the fault degree is ensured to be light. The secondary early warning judgment processing adopts a time window analysis method, and the characteristics of pump efficiency decline or slight faults are required to be detected in three continuous sampling periods, so that false alarm caused by accidental fluctuation is avoided. The judging process also considers the matching relation between the water inflow and the pump efficiency, when the water inflow increases to cause the water pump to deviate from the optimal working point, the pump efficiency naturally drops to be normal, and at the moment, correction judgment is needed according to the characteristic curve of the water pump. The yellow performance degradation warning signal reminds the operator that the equipment performance is degraded but still can continue to operate, and a detailed check is recommended at the next scheduled maintenance, and the signal output contains specific parameters of performance degradation and possible cause analysis.

The three-stage early warning judges the situation that the fault risk of the processing equipment is obviously increased and precautionary measures need to be taken, and the judging condition is that the pump efficiency is in a lower range or medium fault characteristics are detected. The lower pumping efficiency range is defined as between fifty and seventy percent of design efficiency, which indicates that significant problems with the pump are present and equipment damage may be exacerbated. The mid-level fault signature is identified by the mid-level fault probability output by the neural network being higher than zero four, which probability level indicates that the device does have a fault symptom that requires attention. The three-level early warning judgment also needs to analyze the fault development trend, determine whether the fault is worsening by comparing the current fault probability with the historical data, and need to improve the early warning level when the fault probability continuously rises. The judging process considers the coupling effect of a plurality of fault modes, and the comprehensive risk is higher than the simple superposition of single faults when the bearing faults and the impeller wear occur simultaneously. The orange fault risk warning signal requires the operator to closely monitor the equipment status and prepare an emergency plan suggesting that equipment service be scheduled at the appropriate time to avoid further deterioration of the fault. The signal output comprises fault type identification, risk level assessment and recommended maintenance measures, and technical support is provided for maintenance decision.

The four-stage early warning judging and processing equipment faces an emergency situation that serious fault threats need to take action immediately, and judging conditions are that the pump efficiency is lower than a safety threshold or serious fault characteristics are detected. The safety threshold is set at fifty percent of design efficiency below which the pump is substantially disabled from operating properly and continues to operate with equipment damage and personnel safety risks. The severe fault signature is identified by the neural network outputting a probability of severe fault above zero three, which, although not the highest value, has indicated the possibility of severe fault. The four-stage early warning judgment adopts a multiple confirmation mechanism to prevent unnecessary shutdown loss caused by false alarm, at least two independent detection channels are required to confirm the fault state simultaneously, and the detection channels comprise three methods of neural network diagnosis, threshold comparison and trend analysis. The emergency degree of water inflow is also needed to be considered in the judging process, and when the water inflow exceeds the maximum water discharge capacity of the water pump, even if the pump efficiency is normal, four-stage early warning is needed to be triggered to prevent well logging accidents. The red severe fault early warning signal requires that the equipment is immediately stopped and an emergency response program is started, and the signal output ensures timely transmission by adopting a mode of combining audible and visual alarm, short message notification and automatic control. The four-stage warning output signal also contains fault diagnosis reports, emergency treatment suggestions and standby equipment start instructions.

In a specific embodiment, the performing step of performing the coupling judgment matrix construction process based on the fault identification classification result and the real-time pump efficiency data may specifically include the following steps:

carrying out magnitude classification treatment on the current water inflow working conditions according to the real-time water inflow sensor data to obtain four water inflow grade identifiers of low water inflow, medium and high water inflow;

continuously monitoring the current water pump efficiency based on the flowmeter, the pressure transmitter and the power count data to obtain a real-time pump efficiency percentage value;

performing cross mapping treatment on the normal state, slight fault, medium fault and serious fault classification and the water burst grade identification in the fault identification classification result to obtain a fault-water burst quantity association matrix;

performing coupling weight calculation processing on the real-time pump efficiency percentage value and the fault-water inflow quantity association matrix to obtain a pump efficiency-fault coupling coefficient;

and performing matrix element assignment processing based on the pump efficiency-fault coupling coefficient and the water inflow change trend to obtain a water inflow-pump efficiency coupling judgment matrix.

Specifically, the magnitude classification processing of the real-time water inflow sensor data establishes a classification standard based on the underground hydrogeological conditions and the design parameters of the water pump, and the classification process adopts a method of combining threshold judgment and sliding leveling to ensure the stability and accuracy of classification results. The water inflow sensor is deployed in a main water inlet point and a water collecting area in the pit, the water inflow flow change is monitored in real time through the ultrasonic liquid level meter and the electromagnetic flowmeter, and the sensor data are subjected to digital filtering treatment to eliminate the influence of short-term fluctuation. The low water inflow level corresponds to the normal water seepage condition under the well, the water inflow range is zero to fifty cubic meters per hour, and the equipment pressure is smaller but the efficiency is relatively lower when the water pump under the level is operated in a partial load state. The medium-low water inflow level corresponds to a slight water inflow condition, the water inflow range is fifty to one hundred cubic meters per hour, and the efficiency of the water pump starts to enter a more ideal operation interval to be gradually improved. The medium-high water inflow level corresponds to the medium water inflow condition, the water inflow range is one hundred to two hundred cubic meters per hour, and the water pump operates at a higher efficiency near the design working condition but the equipment load is obviously increased. The high water inflow level corresponds to a large water inflow or sudden water inflow situation, the water inflow exceeds two hundred cubic meters per hour, and the efficiency of the water pump is reduced in an overload state and a large fault risk exists.

When at 50+/-5And/h and 100+ -8When the boundary area of/h has the grade overlapping phenomenon, namely when the water inflow is in the overlapping intervals, the system does not switch the grade immediately, and the comprehensive judgment is carried out by combining the water inflow change trend, the duration and the current running state of the equipment. If the water inflow is 45-55Fluctuation in/h interval and rate of change less than 2And when the water inflow is in a significant rising trend in the overlapped interval and the duration exceeds 20 minutes, the system enters a higher level in advance so as to ensure the timeliness of early warning.

In the case of sudden water gushing event, even if the water gushing quantity instantaneously exceeds 200And (h) firstly, verifying the validity of the sensor data by the system, removing sensor faults or external interference factors, and immediately triggering an emergency response mode instead of a conventional high water inflow level processing flow after confirming a real water inflow event. In the equipment start-stop process, abnormal fluctuation of water inflow reading is caused by the hydraulic transient effect, the system pauses grade switching judgment within the first 10 minutes after the water pump start-stop signal is detected, and the pre-warning judgment is carried out by adopting stable working condition data before start-stop. During the maintenance of the sensor, when the main water inflow sensor is offline, the system is automatically switched to a standby sensor, meanwhile, the precision requirement of grading judgment is relaxed, the original strict threshold value is expanded into a fuzzy interval, and the uncertainty of the period is marked in a system log.

The classification treatment adopts a fifteen-minute sliding time window to calculate the average water inflow, avoids frequent grade switching caused by instantaneous fluctuation, and outputs classification results in the form of water inflow grade identification so as to be convenient for subsequent treatment and use.

The continuous monitoring of the water pump efficiency realizes high-precision efficiency measurement based on multi-sensor data fusion and real-time calculation technology, and the relationship among hydraulic power, mechanical power and electric power is comprehensively considered in the monitoring process. The flowmeter adopts an electromagnetic measurement principle, measures the volume flow of the conductive liquid through Faraday electromagnetic induction law, is arranged at the position of the straight pipe section of the water outlet pipeline of the water pump, and the length of the upstream and downstream straight pipe sections meets the requirement of measurement precision, and the measurement precision reaches five percent and the response time is less than one second. The pressure transmitter is respectively arranged at the inlet and outlet positions of the water pump to measure the suction pressure and the discharge pressure, the pressure difference is the actual lift of the water pump, and the pressure transmitter adopts the diffusion silicon technology and has high precision and long-term stability. The power meter measures the input power of the water pump motor, three-phase current and voltage signals are collected through the current transformer and the voltage transformer, and the power calculation considers the influence of the power factor and the harmonic content to ensure the measurement accuracy. The pump efficiency is calculated by adopting a formula of dividing hydraulic power by shaft power, the hydraulic power is calculated by flow, lift and liquid density, and the shaft power is calculated and determined by motor power and transmission efficiency.

When the calculated pump efficiency value is in a critical interval of 75% -80%, the system does not simply divide the fault grades according to the fixed threshold value, but comprehensively evaluates the efficiency change trend, the running condition deviation degree and the historical efficiency base line. If the pump efficiency is within the interval but is stable or rising, and the running time is not more than 120% of the design condition, the system tends to maintain a lower failure risk rating, whereas if the pump efficiency is within an acceptable range but is rapidly falling (the falling rate is more than 1%/hour), the system will increase the failure risk rating and increase the monitoring frequency.

The real-time calculation period is set to be five seconds, the influence of measurement noise is reduced by the calculation result each time through Kalman filtering, and the pump efficiency percentage value is reserved in two decimal places and displayed in a percentage form. The cross mapping of the fault identification classification result and the water burst grade identification establishes a multidimensional association analysis framework, and the mapping process considers the occurrence probability and the influence degree of fault modes under different water burst working conditions. The fault classification result comprises probability distribution of four states, each state corresponds to different equipment health levels and maintenance requirements, and the probability value reflects the confidence of the neural network on each state. The water burst grade identification provides current working condition information, and different water burst grades correspond to different equipment operation conditions and fault risk levels. The cross mapping process adopts a four-by-four matrix structure, the matrix rows correspond to four fault states, the matrix columns correspond to four water burst levels, and each matrix element represents a risk weight corresponding to a combination of the fault states and the water burst levels. The risk weight is determined based on historical fault statistics and expert experience, the lowest weight of the normal state indicates the minimum risk under the condition of low water inflow, and the highest weight of the serious fault indicates the maximum risk under the condition of high water inflow.

When the system detects abnormal combination of normal state and high water inflow, risk assessment is not directly carried out according to a preset matrix, a secondary verification program is started, and the processing records of the sensor calibration state, the confidence interval of the fault recognition algorithm and the similar historical working conditions are checked. If the combination of serious faults and low water inflow is detected, the system takes priority on the possibility of internal faults of the equipment, such as bearing damage or impeller abrasion and other problems which are irrelevant to external working conditions, and the water inflow is low at the moment, so that serious problems of the equipment cannot be covered.

And performing matrix operation on the fault probability vector and the water burst grade mark by mapping calculation to obtain corrected fault probability distribution considering the influence of the working condition. The element values of the fault-water inflow incidence matrix are calculated through a weighted average method, and the weight coefficient is determined according to the historical occurrence frequency and the hazard degree of each combination condition. And (3) calculating the coupling weight of the pump efficiency percentage value and the fault-water inflow incidence matrix to establish a quantitative risk assessment model, wherein the calculation process comprehensively considers the equipment performance index and the fault state information. The pump efficiency percentage value provides a direct measure of the current performance of the device, with lower values indicating a higher risk of failure with poorer device performance. The fault-water inflow correlation matrix provides fault risk assessment considering the influence of working conditions, and the larger the matrix element value is, the higher the risk level of the corresponding combination is. The coupling weight calculation adopts a method of combining fuzzy logic and linear weighting, firstly, the pump efficiency percentage value is converted into a fuzzy set through a membership function, and the membership function describes the relationship between the pump efficiency and the risk level through trapezoidal distribution.

When the pump efficiency value is in a fuzzy boundary region of 60% -65%, the system does not adopt a deterministic membership function value, but introduces an uncertainty region, the membership function presents gradual change characteristics in the region, and meanwhile, the influence of a confidence region (+ -2%) of pump efficiency measurement on final weight calculation is considered. If the pump efficiency change rate is abnormal (the absolute value exceeds 5%/hour), the system automatically expands the fuzzy boundary range from the original + -2.5% to + -5% so as to adapt to the rapid change of the equipment state.

And (3) performing dot multiplication operation on the pump efficiency fuzzy value and the fault-water inflow incidence matrix by a weight calculation formula to obtain a coupling coefficient comprehensively considering the performance and the fault double factors. The calculation process also considers the influence of the pump efficiency change rate, and when the pump efficiency is rapidly reduced, even the current value still needs to be increased in risk weight, and the change rate is determined by calculating the derivative of the pump efficiency time sequence. The numerical range of the pump efficiency-fault coupling coefficient is zero to ten, zero represents no risk and ten represents extremely high risk, and the coefficient value is mapped to the early warning level through a piecewise linear interpolation method.

The matrix element assignment is comprehensively judged based on the pump efficiency-fault coupling coefficient and the water inflow change trend, and the assignment process adopts a method combining dynamic weight adjustment and trend analysis. The water inflow change trend is determined by carrying out linear regression analysis on the water inflow time sequence, and the regression slope reflects the increase and decrease trend and the change rate of the water inflow. The trend analysis uses a thirty minute time window to calculate the rate of change of the water inflow, a positive value indicating an increase in water inflow and a negative value indicating a decrease in water inflow, and the absolute value of the rate of change reflects the severity of the change. The matrix element assignment rule is determined according to a basic value of the pump efficiency-fault coupling coefficient and a correction value of the water inflow change trend, wherein the basic value provides steady state risk assessment, and the correction value considers the influence of dynamic change.

When conflict situations occur that the pump efficiency-fault coupling coefficient and the water inflow change trend point to different risk levels, the system adopts a weighted voting mechanism, and comprehensively arbitrates by combining the equipment operation history, maintenance records and current environmental factors. If the pump efficiency index shows low risk but the water inflow rises sharply, the system takes priority on the urgency of the water inflow change, but marks the inconsistency in the early warning information and recommends an operator to perform manual check. Under extreme working conditions (for example, the water inflow change rate exceeds 50m < 3 >/h), the system pauses the conventional matrix assignment flow, directly enters an emergency early warning mode, and at the moment, all matrix elements are assigned with high risk weight values, so that the system can still keep a conservative safety strategy when data are abnormal or a sensor fails.

When the water inflow is in an ascending trend, the matrix element values need to be adjusted towards a high risk direction, the faster the ascending speed is, the larger the amplitude of adjustment is, and when the water inflow is in a descending trend, the risk level can be properly reduced but the amplitude of reduction is limited. The assignment process also considers the statistical rule of the historical data, and the posterior risk assessment is obtained by combining the prior probability with the current observed value through a Bayesian inference method. The final element value of the water inflow-pump efficiency coupling judgment matrix is determined through comprehensive calculation of all correction factors, and the matrix is in a four-row four-column structure, and each element corresponds to an early warning level of a specific water inflow level and pump efficiency range combination.

The pump room automation monitoring and early warning method in the embodiment of the present application is described above, and the pump room automation monitoring and early warning system in the embodiment of the present application is described below, referring to fig. 2, and one embodiment of the pump room automation monitoring and early warning system in the embodiment of the present application includes:

the acquisition module is used for carrying out multidimensional data acquisition and processing on the underground drainage pump room through the hydraulic vibration sensor, the impeller cavitation acoustic sensor and the bearing temperature rise sensor to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data;

The optimization module is used for carrying out characteristic optimization processing on the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data according to an underground environment self-adaptive variable selection algorithm to obtain a pump efficiency attenuation characteristic vector;

the modeling module is used for carrying out fault mode modeling processing on the pump efficiency attenuation characteristic vector through a hydraulic attenuation model to obtain water pump health state prediction data;

The training module is used for performing deep learning training treatment on the water pump health state prediction data according to the underground water pump fault propagation network to obtain a fault identification classification result;

and the early warning module is used for carrying out grading early warning processing on the operation state of the pump house through water inflow and pump efficiency coupling judgment on the fault identification and classification result to obtain a four-stage early warning output signal.

The pump room automation monitoring and early warning system in the embodiment of the invention is described in detail from the angle of modularized functional entity in fig. 2, and the pump room automation monitoring and early warning device in the embodiment of the invention is described in detail from the angle of hardware processing.

Referring to fig. 3, the embodiment of the invention further provides a pump room automation monitoring and early warning device, where the pump room automation monitoring and early warning device may be a server, and the internal structure of the pump room automation monitoring and early warning device may be as shown in fig. 3. The pump room automation monitoring and early warning device comprises a processor, a memory, a display screen, an input device, a network interface and a database which are connected through a system bus. Wherein the computer is configured to provide computing and control capabilities. The memory of the pump room automation monitoring and early warning device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the pump room automation monitoring and early warning device is used for storing corresponding data in the embodiment. The network interface of the pump room automation monitoring and early warning device is used for communicating with an external terminal through network connection. Which computer program, when being executed by a processor, carries out the above-mentioned method.

It will be appreciated by those skilled in the art that the structure shown in fig. 3 is merely a block diagram of a portion of the structure associated with the present solution and is not intended to limit the pump house automation monitoring and early warning device to which the present solution is applied.

The invention also provides a computer readable storage medium, which can be a nonvolatile computer readable storage medium, and can also be a volatile computer readable storage medium, wherein the computer readable storage medium stores instructions, and when the instructions run on a computer, the instructions cause the computer to execute the steps of the pump house automation monitoring and early warning method.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or partly in the form of a software product, or all or part of the technical solution, which is stored in a storage medium, and includes several instructions for causing a pump house automation monitoring and early warning device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. The storage medium includes a U disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.

The foregoing embodiments are merely for illustrating the technical solution of the present invention, but not for limiting the same, and although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that modifications may be made to the technical solution described in the foregoing embodiments or equivalents may be substituted for parts of the technical features thereof, and that such modifications or substitutions do not depart from the spirit and scope of the technical solution of the embodiments of the present invention in essence.

Claims (8)

1. A pump room automation monitoring and early warning method, characterized in that the method comprises: Multi-dimensional data collection and processing of underground drainage pump room is carried out through hydraulic vibration sensors, impeller cavitation acoustic sensors and bearing temperature rise sensors to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data; According to the downhole environment adaptive variable selection algorithm, the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data are subjected to feature optimization processing to obtain a pump efficiency attenuation feature vector, including: dividing the downhole water inflow condition into four levels of low flow, medium-low flow, medium-high flow and high flow based on the water inflow classification matrix to obtain a water inflow condition classification label; extracting key inflection point parameters of the pump head-flow characteristic curve according to the water inflow condition classification label to obtain the best efficiency point flow, shut-off head, maximum flow and efficiency decline inflection point parameters; inputting the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data into the principal component analysis algorithm for correlation weight calculation processing to obtain a sensor parameter weight matrix; performing dynamic correction processing on the sensor parameter weight matrix for changes in downhole temperature, humidity and pH value to obtain an environmental adaptive weight coefficient; screening and optimizing the core characteristic variables based on the environmental adaptive weight coefficient to obtain a pump efficiency attenuation feature vector; The pump efficiency attenuation characteristic vector is subjected to fault mode modeling through a hydraulic attenuation model to obtain water pump health status prediction data, including: calculating the cavitation coefficient based on the cavitation acoustic emission intensity and impeller cavitation acoustic sensor data in the pump efficiency attenuation characteristic vector to obtain the cavitation loss coefficient; calculating the wear loss coefficient based on the water quality pH value and particle concentration parameters in the pump efficiency attenuation characteristic vector to obtain the impeller wear loss coefficient; inputting the cavitation loss coefficient and the impeller wear loss coefficient into the water pump similarity law to perform pump efficiency attenuation law modeling to obtain pump efficiency time-varying attenuation data; performing polynomial fitting calculation on the pump efficiency time-varying attenuation data to obtain water pump head-flow characteristic curve degradation parameters; performing comprehensive health status evaluation based on the water pump head-flow characteristic curve degradation parameters and bearing water lubrication theory to obtain water pump health status prediction data; Performing deep learning training on the water pump health status prediction data based on the downhole water pump fault propagation network to obtain a fault identification and classification result; The fault identification and classification results are used to perform graded early warning processing on the pump room operation status through water inflow-pump efficiency coupling judgment, and a four-level early warning output signal is obtained. 2. The pump room automation monitoring and early warning method according to claim 1 is characterized in that the multi-dimensional data acquisition and processing of the underground drainage pump room is performed by a hydraulic vibration sensor, an impeller cavitation acoustic sensor and a bearing temperature rise sensor to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data, including: The hydraulic vibration sensor collects and processes signals from the piezoelectric vibration detection unit and the water pressure pulsation measurement unit at the water pump inlet to obtain hydraulic excitation spectrum data in the 0.1Hz-5kHz frequency band. Performing amplitude extraction processing on the hydraulic excitation spectrum data to obtain the hydraulic pulsation amplitude; performing broadband acoustic emission detection processing on the 20kHz-200kHz frequency band of the impeller housing using the impeller cavitation acoustic sensor to obtain the bubble burst acoustic characteristic signal; Performing acoustic emission intensity calculation processing on the bubble collapse acoustic characteristic signal to obtain cavitation acoustic emission intensity; The bearing seat temperature change is monitored and processed with a resolution of 0.1°C based on the bearing temperature rise sensor to obtain the bearing temperature rise data. 3. The pump room automation monitoring and early warning method according to claim 1 is characterized in that the deep learning training processing of the water pump health status prediction data based on the downhole water pump fault propagation network to obtain the fault identification and classification results includes: Inputting the water pump health status prediction data into the input layer of the multi-layer neural network for data reception and processing, and obtaining input data corresponding to the impeller cavitation acoustic emission intensity, bearing temperature rise rate, and hydraulic vibration main frequency; Performing ReLU activation function calculation processing on the hidden layer neuron nodes based on the input data to obtain a nonlinear feature mapping result; According to the nonlinear feature mapping results, the four nodes in the output layer are classified into fault states to obtain the classification probabilities of normal state, minor fault, moderate fault and major fault; Inputting the classification probability into the cross entropy loss function and the mean square error loss function to perform weighted combination loss calculation processing to obtain the network training loss value; The network parameters are back-propagated and optimized based on the network training loss value to obtain a fault identification and classification result. 4. The pump room automation monitoring and early warning method according to claim 1 is characterized in that the fault identification classification result is judged by water inflow-pump efficiency coupling to carry out hierarchical early warning processing on the pump room operating state to obtain a four-level early warning output signal, including: A coupling judgment matrix is constructed based on the fault identification and classification results and the real-time pump efficiency data to obtain a water inflow-pump efficiency coupling judgment matrix; According to the water inflow-pump efficiency coupling judgment matrix, a first-level warning judgment process is performed when the pump efficiency is higher than the set threshold and the fault is identified as a normal state, and a green normal operation warning signal is obtained; Inputting the pump efficiency in the medium range or detecting a minor fault feature in the water inflow-pump efficiency coupling judgment matrix into the secondary warning judgment for processing to obtain a yellow performance degradation warning signal; A three-level warning judgment process is performed on the situation where the pump efficiency in the water inflow-pump efficiency coupling judgment matrix is in a low range or a medium fault feature is detected, and an orange fault risk warning signal is obtained; Based on the situation in the water inflow-pump efficiency coupling judgment matrix where the pump efficiency is lower than the safety threshold or a serious fault feature is detected, a four-level warning judgment process is performed to obtain a red serious fault warning signal as a four-level warning output signal. 5. pump room automation monitoring and early warning method according to claim 4, is characterized in that, described coupling judgment matrix construction process is carried out based on described fault identification classification result and real-time pump efficiency data, obtains water inflow-pump efficiency coupling judgment matrix, including: The current water inflow condition is classified according to the real-time water inflow sensor data, and four water inflow grade identifications are obtained: low water inflow, medium-low water inflow, medium-high water inflow and high water inflow; Continuously monitor the current water pump efficiency based on the flow meter, pressure transmitter and power meter data to obtain the real-time pump efficiency percentage value; Cross-mapping the normal state, minor fault, moderate fault and severe fault classification in the fault identification classification result with the water inflow level identifier to obtain a fault-water inflow association matrix; Performing coupling weight calculation processing on the real-time pump efficiency percentage value and the fault-water inflow association matrix to obtain a pump efficiency-fault coupling coefficient; Matrix element assignment processing is performed based on the pump efficiency-fault coupling coefficient and the water inflow variation trend to obtain a water inflow pump efficiency coupling judgment matrix. 6. A pump room automation monitoring and early warning system, characterized in that it is used to implement the pump room automation monitoring and early warning method according to any one of claims 1 to 5, and the pump room automation monitoring and early warning system comprises: The acquisition module is used to collect and process multi-dimensional data of the underground drainage pump room through hydraulic vibration sensors, impeller cavitation acoustic sensors and bearing temperature rise sensors to obtain hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data; An optimization module is used to perform feature optimization processing on the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data according to a downhole environment adaptive variable selection algorithm to obtain a pump efficiency attenuation feature vector, including: classifying the downhole water inflow condition into four levels of low flow, medium-low flow, medium-high flow and high flow based on a water inflow classification matrix to obtain a water inflow condition classification label; extracting key inflection point parameters of the pump head-flow characteristic curve according to the water inflow condition classification label to obtain the best efficiency point flow, shut-off head, maximum flow and efficiency decline inflection point parameters; inputting the hydraulic pulsation amplitude, cavitation acoustic emission intensity and bearing temperature rise data into a principal component analysis algorithm for correlation weight calculation processing to obtain a sensor parameter weight matrix; performing dynamic correction processing on the sensor parameter weight matrix for changes in downhole temperature, humidity and pH value to obtain an environmental adaptive weight coefficient; and screening and optimizing the core characteristic variables based on the environmental adaptive weight coefficient to obtain a pump efficiency attenuation feature vector. A modeling module is used to perform fault mode modeling processing on the pump efficiency attenuation characteristic vector through a hydraulic attenuation model to obtain water pump health status prediction data, including: performing cavitation coefficient calculation processing based on the cavitation acoustic emission intensity and impeller cavitation acoustic sensor data in the pump efficiency attenuation characteristic vector to obtain a cavitation loss coefficient; performing wear loss coefficient calculation processing based on the water quality pH value and particle concentration parameters in the pump efficiency attenuation characteristic vector to obtain an impeller wear loss coefficient; inputting the cavitation loss coefficient and the impeller wear loss coefficient into the water pump similarity law to perform pump efficiency attenuation law modeling processing to obtain pump efficiency time-varying attenuation data; performing polynomial fitting calculation processing on the pump efficiency time-varying attenuation data to obtain water pump head-flow characteristic curve degradation parameters; performing comprehensive health status evaluation processing based on the water pump head-flow characteristic curve degradation parameters and bearing water lubrication theory to obtain water pump health status prediction data; A training module is used to perform deep learning training on the water pump health status prediction data according to the downhole water pump fault propagation network to obtain a fault identification and classification result; The early warning module is used to perform graded early warning processing on the pump room operation status based on the fault identification and classification results through water inflow-pump efficiency coupling judgment, and obtain a four-level early warning output signal. 7. A pump room automatic monitoring and early warning device, characterized in that it includes a memory and a processor, the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the pump room automatic monitoring and early warning method described in any one of claims 1 to 5. 8. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the pump room automatic monitoring and early warning method according to any one of claims 1 to 5.