TWI787955B - Method and system for detecting abnormal temperature of device online - Google Patents

Abstract

A method and system for detecting abnormal temperature of device online are provided. The method includes: receiving a plurality of historical data of a device to be inspected. The historical data is used as a training set, and the training set is input to a prediction model. A number of prediction data and a threshold value are input to the trained prediction model, thereby obtaining a plurality of predicted temperature data and a prediction result.

Description

Online detection method and system for equipment temperature abnormality

The present invention relates to a service life prediction method, and in particular to an online detection method and system for abnormal temperature of equipment.

In an industrial environment, the operation of electronic equipment and the maintenance of environmental parameters have a great impact on the preservation of production lines and instruments. Nowadays, regular maintenance is usually used to reduce sudden failures. However, the effect of regular maintenance to prevent sudden failures is still not effective enough.

Industrial air-conditioning systems often have sudden failures after long-term use. However, if the air-conditioning system cannot be repaired and replaced in time, the temperature of the industrial environment will not be maintained within a set value, resulting in damage to equipment and product productivity. Impact. Existing equipment diagnosis or detection systems are for diagnosing current equipment conditions, but have no function of predicting operating parameters and abnormal time points, and do not provide data or graphics that allow operators to quickly understand equipment conditions.

The present invention provides an online detection method and system for abnormal temperature of equipment, and a prediction model is trained by using multiple historical data of the operating temperature of the equipment to be tested. In addition, the boundary set value and the number of forecast data are input into the forecast model, so as to obtain the forecast result of the future operating temperature of the equipment, so as to provide enterprises with early implementation of preventive, repair or maintenance measures before equipment abnormalities or failures occur.

The online detection method for device temperature abnormality of the present invention includes: receiving multiple historical data of the device to be tested; using the historical data of the device to be tested as a training set; training a prediction model according to the training set, wherein the prediction model is based on a machine The learning algorithm is trained; the number of forecast data and the set value of the boundary are input into the forecast model to obtain multiple forecast temperature values and forecast results of the device under test.

In an embodiment of the present invention, the aforementioned boundary set value includes a temperature upper limit set value and a temperature lower limit set value set by a user.

In an embodiment of the present invention, the above-mentioned step of inputting the number of forecast data and the boundary setting value into the forecast model to obtain the forecast temperature value of the device under test and the forecast result further includes: the forecast The number of data items and the boundary set value are input into the prediction model to obtain the predicted temperature value; judge whether the predicted temperature value exceeds the boundary set value to obtain the judgment result; wherein if any of the predicted temperature values exceeds the For the boundary set value, the exceeded predicted temperature value, the time corresponding to the exceeded predicted temperature value, and the warning message will be used as the judgment result; if the predicted temperature values are all between the boundary set value, the safety message will be used as The judgment result.

In an embodiment of the present invention, the above-mentioned online detection method for abnormal temperature of the equipment further includes: setting the number of predicted steps as N, and setting each step as 0 hours; and setting N times in the future after receiving the boundary set value point. After judging whether the predicted temperature value exceeds the boundary set value, it further includes: if the predicted temperature values are all between the boundary set value, then after judging that the boundary set value has been received after N times O hours The N time points of N time points are used to obtain the estimated abnormal time points by using the prediction model trained based on the machine learning model and the historical data of the equipment operating temperature.

In an embodiment of the present invention, the above-mentioned predicted temperature values include a time point and a temperature value at each predicted temperature value.

In an embodiment of the present invention, after the above-mentioned step of judging whether the predicted temperature value exceeds the boundary set value and obtaining the judgment result, it includes: drawing the operating temperature prediction of the device under test according to the predicted temperature value and the boundary set value Graphics; output the running temperature prediction graphics.

The online detection system for abnormal device temperature in the embodiment of the present invention includes (but not limited to) storage and processor. The memory stores multiple pieces of equipment temperature data. The processor is configured to receive multiple pieces of historical data of the device under test, and use the historical data of the device under test as a training set. A predictive model is trained on the training set. The number of forecast data and the boundary setting value are input into the forecast model to obtain multiple forecast temperature values and forecast results of the device under test. The predictive model is trained based on machine learning algorithms.

Based on the above, this disclosure proposes an online detection method and system for equipment temperature abnormalities, which can use machine learning models and a rolling multi-step prediction method to predict the time point of abnormal temperature of the equipment and the time point of future failures. Provide enterprises with early detection or maintenance before equipment failure or abnormal operation to prevent failures.

When the equipment has been used for a period of time, the equipment will malfunction or the temperature will be abnormal, and the historical data of the equipment’s operating temperature detected by the sensors in the equipment cannot allow the operator or operator to directly judge the equipment. Therefore, it is necessary to use a prediction method to predict the operating status of the equipment and the time point when an abnormality may occur, so that the operator can perform fault-preventive maintenance actions or check and repair in advance. Accordingly, the disclosure provides an online detection method and system for abnormal temperature of equipment, which can realize the prediction of abnormal temperature online.

FIG. 1 is a schematic diagram of an online detection system for abnormal temperature of equipment according to an embodiment of the present invention. In the following embodiments, the online detection method and system for abnormal device temperature can be realized by the electronic device 1 with computing functions. The electronic device 1 includes a processor 110 , a storage 120 and a communication transceiver 130 . The processor 110 may be hardware (such as chipset, processor, etc.), software components (such as operating system, application programs, etc.) capable of computing and processing, or a combination of hardware and software components. The processor 110 is, for example, a central processing unit (Central processing unit, CPU), a graphics processing unit (Graphics processing unit, GPU), or other programmable microprocessor (Microprocessor), digital signal processor (Digital signal processor, DSP), programmable controller, Application specific integrated circuits (ASIC), programmable logic device (Programmable logic device, PLD) or other similar devices.

The storage 120 is coupled to the processor. The storage 120 is, for example, any type of fixed or removable random access memory, read-only memory, flash memory, secure digital card, hard disk, or other similar devices or a combination of these devices. At least one program code segment, at least one local database, and at least one data table are stored in the storage 120, and the above-mentioned program code segment is executed by the processor 110 after being installed to realize the online detection method and system for device temperature abnormality .

The communication transceiver 130 is coupled to the processor 110 . The communication transceiver 130 is, for example, a transceiver supporting wired networks such as Ethernet (Ethernet), an optical fiber network, or a cable (which may include (but not limited to) components such as connection interfaces, signal converters, and communication protocol processing chips. ), or a transceiver (which may include (but is not limited to) antennas, digital to analog/analog-to-digital converters, protocol processing chips, etc.). In one embodiment, the communication transceiver 130 is used to transmit or receive data.

FIG. 2 is a flowchart of an online detection method for abnormal temperature of equipment according to an embodiment of the present invention. In step S210, multiple pieces of historical data of the device under test are received. Specifically, the online detection method for equipment temperature abnormality of the present invention detects and records multiple temperature values during the operation of the equipment through the temperature sensor installed in the equipment. And, thereby collecting historical data of equipment such as operating temperature value, operating time and/or equipment status.

Fig. 3 is a schematic diagram of an online detection system for abnormal temperature of equipment according to an embodiment of the present invention. In this embodiment, the local database DL stores multiple pieces of historical data. Wherein, the multiple pieces of historical data are temperature time series data, and the formed data table is stored in the local database DL.

In one embodiment, the historical data of the device is stored in a local database or storage. Accordingly, in step S210, the processor 110 receives and reads multiple pieces of historical data from the local database DL. Moreover, after the processor 110 receives multiple pieces of historical data of the device under test, it performs data pre-processing on the historical data through the data pre-processing program SP. For example, assume that the original sensor collects 12,730 pieces of historical data, and then the data is converted into the same unit of hours or seconds by taking an average. Next, perform data preprocessing. In one embodiment, the aforementioned data pre-processing is to smooth the original data by using Locally estimated scatterplot smoothing (LOESS).

The pre-processing of the above data can also be normalization (Normalization) processing of the temperature values (that is, historical data) of multiple pieces of equipment during operation. For example, the maximum value among the multiple historical data is converted to 1, the minimum value among the historical data is converted to 0, and the remaining values are distributed between 0 and 1 according to the proportion. Accordingly, the value of the historical data is set between 0 and 1. For example, the above-mentioned normalization processing can be performed through MinMaxScaler, but in practical applications, other data pre-processing methods can also be used, so this case should not be limited to this.

In step S220, the historical data of the device under test is used as a training set. In this step S220, the processor 110 uses the historical data that has completed data pre-processing as a training set. Next, in step S230, the prediction model is trained according to the training set. Specifically, the processor 110 uses the training set as an input parameter of the predictive model to train the predictive model program PMP, and exports it to the web application framework WPF. In the present invention, the predictive model is trained based on a machine learning algorithm.

In step S240, the processor 110 inputs the number of forecast data and the boundary setting value into the forecast model to obtain multiple forecast temperature values and forecast results of the device under test. The number of predicted temperature values corresponds to the number of predicted data. In one embodiment, the online detection system for abnormal device temperature is connected to the web application program WAP through the web application framework. In this embodiment, the web application framework WPF is a flask program, a program written by the Python programming language, and the window program WA connected to the local database is a Python programming language that has the function of reading/accessing data. Application programming interface. For example, the window program WA connected to the local database can be Java Database Connectivity (JDBC) or Open Database Connectivity (ODBC), etc., so the present invention should not be limited thereto. As shown in Figure 3, the prediction model program is exported to the web application framework WPF after being trained. Then, the user enters the boundary setting value and the number of predicted data to be predicted in the web application program WAP (for example, the user inputs 10 data, that is, the predicted temperature value of the next 10 data will be output). Accordingly, the web application framework WPF, which communicates with the web application program WAP, inputs the boundary setting value and the number of predicted data items into the above-mentioned trained forecasting model, so as to predict the corresponding number of items based on the number of predicted data items predicted temperature value. It should be added that the system architecture diagram in Fig. 3 provides an example architecture for those in this field, and the features of the online detection method and system of the present invention should be achieved by other architectures or programs, so the present invention does not include This example should be limited.

In one embodiment, the boundary setting value includes a temperature upper limit setting value and a temperature lower limit setting value set by the user. For example, the user inputs the temperature upper limit setting value and the temperature lower limit setting value of the device under test through the web application program WAP. It is supplemented that the upper temperature limit setting value and the temperature lower limit setting value can be the recommended operating temperature in the operation manual of the device or the operating temperature according to the operating experience of the user. Then, the network application program WAP returns the boundary setting value input by the user and the future number of items that the user wants to predict (that is, the number of predicted data items) to the web application framework WPF, and stores them in the data table D and local data Among the library DL.

FIG. 4 is a flow chart of obtaining a predicted temperature value according to an embodiment of the present invention. In step S240, the online detection method for equipment temperature abnormality of the present invention further includes: inputting the number of prediction data and the boundary setting value into the above-mentioned trained prediction model to obtain the predicted temperature value (step S241). That is to say, the processor 110 inputs the predicted data ratio and the threshold setting value into the trained prediction model, so as to calculate the predicted temperature value through the prediction model. Next, in step S242, it is judged whether the predicted temperature value exceeds the threshold setting value to obtain a judgment result, and the judgment result is output. Specifically, the processor 110 compares and judges the predicted temperature value and the boundary setting value set by the user. If any of the predicted temperature values exceeds the boundary set value, the exceeded predicted temperature value, the time corresponding to the exceeded predicted temperature value, and the warning message are taken as the judgment result. On the contrary, if the predicted temperature values are all between the boundary set values, the safety message is taken as the judgment result. In other words, the webpage application framework WPF that stores the trained forecasting model program PMP inputs the boundary setting value and the number of predicted temperature items input by the aforementioned user into the forecasting model program PMP, and then obtains the forecast corresponding to the number of predicted temperature items Temperature value and judgment result. On the other hand, the processor 110 can output the judgment result to the user's electronic device or the web application program WAP through the communication transceiver 130 to be displayed on the display of the corresponding electronic device or web application program WAP.

FIG. 5 is a graph showing a forecasted operating temperature according to an embodiment of the present invention. In another embodiment, the online detection method for equipment temperature abnormality of the present invention further includes drawing the operation temperature prediction graph of the equipment under test, so that the operator can more easily grasp and understand the operation status of the equipment under test, and make the Operators can more efficiently read the predicted temperature value of the device under test, thereby improving detection efficiency and judgment accuracy. After the above-mentioned step of judging whether the predicted temperature value exceeds the boundary set value to obtain the judgment result (step S242), the online detection method for device temperature abnormality of the present invention further includes: the processor draws a graph according to the predicted temperature value and the boundary set value. The operating temperature prediction graph of the device under test (step S243), and the processor outputs the operating temperature prediction graph (step S244). If any of the predicted temperature values exceeds the boundary set value, the exceeded predicted temperature value and the time corresponding to the exceeded predicted temperature value are marked on the operating temperature predicted graph. Accordingly, if the predicted temperature value is between the boundary set value (that is, the temperature upper and lower threshold values input by the user), the web application page will display a safety signal (for example, remind the user with a green light), Let the user know clearly and quickly that the predicted temperature values are all within the normal range. Conversely, if the predicted temperature value is greater than the upper temperature limit of the boundary set value or lower than the lower temperature limit of the boundary set value, the web application will display a warning signal (for example, remind the user with a red light). The warning signal also includes the predicted temperature value and the time point of exceeding the critical value of the boundary, so that the operator can detect and repair the equipment in advance, so as to prevent the sudden failure of the equipment during operation, and avoid the failure of the equipment from causing damage to the production line. Loss of production capacity.

In one embodiment, before the above-mentioned step of receiving the historical data of the device under test (step S210), the online detection method for device temperature abnormality of the present invention further includes: detecting and recording multiple current operating temperatures of the device under test , and use the current operating temperature as historical data. The current operating temperature is obtained by at least one sensor detecting multiple temperature values of the device under test.

In addition, a rolling forecast can be further added to improve the accuracy of the forecast and provide users with diversified operations. The method for online detection of abnormal device temperature of the present invention further includes setting the number of prediction steps (Prediction step) as N, and setting each step as O hours. N and O are positive integers. Next, N time points in the future after receiving the boundary setting value are set. Here, time is used as an independent variable (Independent variable) to predict the dependent variable (Controlled variable) of future data, that is, the predicted temperature value. For example, with N=3, O=1, and the received boundary set value is the 110th hour, it is predicted that the prediction parameters at three future time points after receiving the boundary set value, that is, the 111th, 111th, Predicted temperature values for 112 and 113 hours. In another embodiment, the above-mentioned received boundary setting value may also be a judgment time point set by the user.

Next, it is judged whether the predicted temperature value exceeds the threshold setting value. If the predicted temperature values are all between the boundary set values, the predicted temperature values at the next N time points are repeatedly predicted. Continue the rolling forecast until the predicted temperature value exceeds the boundary set value and stop the rolling forecast. That is to say, by taking N=3, O=1 and receiving the boundary set value at the 110th hour, reset the next 3 time points after receiving the boundary set value after 3×1 hour (i.e. 114, 115, 116 Hour). For example, as shown in Figure 5, the temperature forecast graph for the next 10 time points (that is, 111 to 120 hours), if the upper limit temperature value of the boundary set value is 30, and the lower limit set value is 10, then the temperature forecast graph will exceed the upper limit. The temperature value of the 116th hour of the temperature value 30 and the corresponding time (116th hour) are sent to the user's device through the web application or through the communication transceiver to notify the user of the warning message and the temperature value exceeding the limit set value (32°C) and its corresponding time (116th hour). It should be added that the unit of time mentioned in the present invention can be set as seconds, hours, seconds, or hours according to the actual situation of the user, so this case should not be limited thereto.

If the loop stops, that is, one of the predicted temperature values exceeds the boundary set value, the time point corresponding to the predicted temperature value exceeding the boundary set value is used as the predicted abnormal time point. Accordingly, through the online detection method and system for abnormal device temperature of the present invention, the user can predict and draw the predicted operating temperature value of the device under test after a certain time point, thereby achieving efficient control of the device's operating status and Effect of operating temperature. In this way, the operator detects and repairs the equipment in advance according to the predicted temperature value and the operating temperature forecast graph, so as to prevent the failure of the equipment.

In the present invention, the machine learning model adopts a support vector regression (SVR) algorithm. For example, the predictive model of the present invention can be based on machine learning algorithms (for example, support vector regression (Support vector regression, SVR), convolutional neural network (Convolutional Neural Network, CNN), deep neural network (Deep Neural Network) , DNN), recurrent neural networks (recurrent neural networks, RNN), long short-term memory (Long Short-Term Memory, LSTM), or other algorithms) training. Machine learning algorithms can analyze training samples to obtain patterns from them, so as to predict unknown data through patterns. The prediction model is a machine learning model constructed after learning, and inferences are made based on the evaluation data. Moreover, the above-mentioned rolling forecast can establish EWMA statistics through the (Exponentially Weighted Moving Average, EWMA) control chart to smooth and monitor the equipment temperature.

To sum up, in the online detection method and system for abnormal equipment temperature in the embodiment of the present invention, by means of a machine learning algorithm, the machine learning algorithm is used to train the boundary setting values (that is, the normal operating temperature values of different equipment) that meet the various equipment. Prediction model, so as to provide users with more accurate and suitable predicted temperature values for different devices.

Although the present invention has been disclosed above with the embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the technical field may make some changes and modifications without departing from the spirit and scope of the present invention. The scope of protection of the present invention should be defined by the scope of the appended patent application.

1: Electronic device 110: Processor 120: storage 130: Communication transceiver S210～S240, S241～S244: steps DL: local database WA: Windows program to connect to local database SP: data preprocessing program PMP: Predictive Modeling Program D: data sheet WAP: Web Application WPF: Web Application Framework

FIG. 1 is a schematic diagram of an online detection system for abnormal temperature of equipment according to an embodiment of the present invention. FIG. 2 is a flowchart of an online detection method for abnormal temperature of equipment according to an embodiment of the present invention. Fig. 3 is a schematic diagram of an online detection system for abnormal temperature of equipment according to an embodiment of the present invention. FIG. 4 is a flow chart of obtaining a predicted temperature value according to an embodiment of the present invention. FIG. 5 is a graph showing a forecasted operating temperature according to an embodiment of the present invention.

S210~S240: steps

Claims (9)

An online detection method for abnormal temperature of equipment, comprising: receiving a plurality of historical data of a device under test; using the historical data of the device under test as a training set; training a prediction model according to the training set, wherein the prediction model It is trained based on a machine learning algorithm; and a number of prediction data and a boundary setting value are input into the prediction model to obtain multiple prediction temperature values and a prediction result of the device under test; wherein the prediction temperature The number of values corresponds to the number of prediction data; wherein the number of prediction data and the boundary setting value are input into the prediction model to obtain the predicted temperature values of the device under test and the steps of the prediction results include: Inputting the number of forecast data and the boundary setting value into the forecasting model to obtain the predicted temperature values; and judging whether the predicted temperature values exceed the boundary setting value to obtain a judgment result; After the step of predicting whether the temperature value exceeds the boundary set value and obtaining a judgment result, it includes: drawing an operating temperature prediction graph of the device under test according to the predicted temperature values and the boundary set value; outputting the operating temperature prediction graph; wherein if any of the predicted temperature values exceeds the limit set If the value is fixed, the exceeding predicted temperature values and the time corresponding to the exceeding predicted temperature value are marked on the operating temperature prediction graph. The online detection method for equipment temperature abnormality as described in Claim 1, wherein the boundary set value includes a temperature upper limit set value and a temperature lower limit set value set by a user. The online detection method for equipment temperature abnormality as described in claim item 1, wherein if any of the predicted temperature values exceeds the boundary set value, the exceeded predicted temperature values, the exceeded predicted The time corresponding to the temperature value and a warning message are used as the judgment result; wherein if the predicted temperature values are all between the boundary set values, a safety message is used as the judgment result. The online detection method for device temperature abnormality as described in claim 1, wherein before the step of receiving the historical data of the device under test, it includes: detecting and recording multiple current operating temperatures of the device under test, and The current operating temperatures are used as the historical data, wherein the current operating temperatures are obtained by at least one sensor detecting a plurality of temperature values of the device under test. The online detection method for device temperature abnormality according to claim 1, wherein the predicted temperature values include a time point and a temperature value on each of the predicted temperature values. The online detection method of equipment temperature abnormality as described in claim item 1 further includes: setting a prediction step number as N, and setting each step as O hours; Set N time points in the future after receiving the boundary set value; wherein, after judging whether the predicted temperature values exceed the boundary set value, it further includes: if the predicted temperature values are all within the boundary set value In between, it is judged that N time points after N times O hours have elapsed since receiving the boundary setting value, so as to obtain an estimated abnormal time point. An online detection system for abnormal temperature of equipment, comprising: a storage device storing a plurality of historical data of a device under test; a processor configured to: receive the historical data; use the historical data of the device under test The data is used as a training set; a prediction model is trained according to the training set, wherein the prediction model is trained based on a machine learning algorithm; a number of prediction data and a boundary setting value are input into the prediction model to obtain the expected The number of predicted temperature values of the measuring equipment and a predicted result; wherein the number of predicted temperature values corresponds to the number of predicted data; wherein the number of predicted data and the boundary setting value are input into the prediction model to obtain In the steps of the predicted temperature values of the device under test and the predicted results, the processor is further configured to: input the number of predicted data and the set value of the boundary into the predicted model to obtain the predicted values temperature value; and judging whether the predicted temperature values exceed the boundary set value to obtain a judgment Result; wherein after the step of judging whether the predicted temperature values exceed the boundary set value to obtain a judgment result, the processor is further configured to: draw the expected temperature value and the boundary set value according to the predicted temperature values and the boundary set value An operating temperature prediction graph of the measuring equipment; output the operating temperature prediction graph; wherein if any of the predicted temperature values exceeds the boundary set value, the exceeded predicted temperature values and the exceeded The time mark corresponding to the predicted temperature value is on the running temperature predicted graph. The online detection system for abnormal temperature of equipment as described in claim 7, wherein the boundary set value includes a temperature upper limit set value and a temperature lower limit set value set by the user. The online detection system for abnormal equipment temperature as described in claim item 7, wherein if any of the predicted temperature values exceeds the boundary set value, the exceeded predicted temperature values, the exceeded predicted The time corresponding to the temperature value and a warning message are used as the judgment result; wherein if the predicted temperature values are all between the boundary set values, a safety message is used as the judgment result.

Images