CN1515382A - On-line Intelligent Control System of Machine Tool Cutting Chatter - Google Patents

Abstract

An online intelligent control system for machine tool cutting chatter is suitable for the field of precision flexible manufacturing and processing. It is characterized in that the online adjustment of the dynamic characteristics of the processing system is mainly carried out by means of a mechanical structure containing electrorheological materials, and the method designs an intelligent boring bar whose dynamic characteristics can be quickly adjusted online by using electrorheological materials for the boring system. By applying different electric field intensities ranging from 0 to 2000 volts/mm to the electrorheological material, the natural frequency of the boring bar can be made to have a variation range of 30 Hz. The present invention can quickly predict the omen of cutting chatter according to the cutting vibration signal, and adjust the dynamic characteristics of the cutting system online in time according to the information contained in the acceleration signal, so as to suppress the cutting chatter in its budding state, without producing machining vibration marks on the surface of the workpiece, and without affecting The normal progress of the cutting process ensures the quality and efficiency of processing.

Description

On-line Intelligent Control System of Machine Tool Cutting Chatter

The online intelligent control system of machine tool cutting chatter of the present invention is a divisional application of the application titled "On-line Intelligent Control Method and System of Machine Tool Cutting Chatter" and the filing date of this case: November 19, 2001, application number : 01144486.x.

technical field

The on-line intelligent control method and system of machine tool cutting chatter belong to the field of mechanical manufacturing automation and intelligent manufacturing. life. This technology is suitable for precision flexible manufacturing and processing systems with a high degree of automation, and can be widely used in industries such as aviation, aerospace and automobiles that require precision processing.

Background technique

Cutting chatter belongs to self-excited vibration, which is a very strong relative vibration between the tool and the workpiece generated by the metal cutting machine tool during the cutting process. Its causes, occurrence and development rules are inherently related to the cutting process itself and the dynamic characteristics of metal cutting machine tools. There are many influencing factors, and it is a very complex mechanical vibration phenomenon. With the development of factory automation, the flexibility of mechanical processing requires that cutting can be performed on different workpieces and under different working conditions. Therefore, the method of chatter prevention and control often cannot fundamentally eliminate the occurrence of chatter. On-line monitoring, forecasting and control has become a key technology to improve the stability of the cutting system.

Due to the suddenness and uncertainty of chatter in the cutting system, the time course from normal cutting to chatter is very short, generally within a few hundred milliseconds. So online monitoring and control of cutting chatter is very difficult. At present, the more successful methods can be classified into two categories: one uses the method of systematic modeling of the cutting system for chatter control, and the other uses online adjustment of cutting processing parameters (spindle speed, feed rate, cutting depth, etc.) to suppress chatter. However, due to the complexity of the cutting processing system, it is very difficult to establish an accurate mathematical model of the system. At the same time, the serious lag of the response of the cutting processing machinery system makes the above two types of methods unable to control the cutting stability well on-line, and cannot completely control chatter. The phenomenon is eliminated.

Contents of the invention

The main purpose of the present invention is to develop a general online control method of cutting chatter without affecting the flexibility of the manufacturing unit for the flexible manufacturing unit, so as to improve the processing quality and production efficiency. The present invention mainly aims at overcoming the obstacles that were difficult to break through in the previous on-line chatter control technology, and uses intelligent materials to collect cutting vibration information from sensors to regulate the dynamic characteristics of mechanical structures to realize online control of cutting chatter.

The design concept of the present invention is based on the theory of controlling the time-varying dynamic characteristics of the cutting system to improve the cutting stability as its theoretical basis, and the original machine tool cutting system is slightly changed during the specific implementation. The original cutting system is mainly composed of three parts: machine tool, workpiece and cutting tool system. The present invention only adopts the embedded intelligent material in the design of the key parts (such as boring bar, tool handle, etc.) that affect the stability of the cutting system. The method uses the characteristics that the dynamic characteristics of intelligent materials can be adjusted quickly in real time, and the dynamic characteristics of the entire processing system are adjusted online and quickly, eliminating the conditions for chatter to avoid the occurrence of cutting chatter.

The control method designed in the technical solution of the present invention is partly referred to in flow chart 1, and CDM online recognition flutter prediction is referred to CDM information flow chart 3 and CDM online recognition flutter omen flow chart 2, and the method is characterized in that it includes the following steps in sequence :

(1) System initialization, setting: initialization electric field strength E0,

Sampling frequency Fs;

(2) Flutter recognition: input the acceleration signal into the flutter recognition program module CDM, and perform circular discrimination and operation until the flutter omen is recognized;

(3) Flutter control:

(a) Once the flutter omen is identified, calculate the flutter frequency F _C at this time, let: F _post = F _C , F _post : flutter frequency before applying the electric field strength;

(b) Apply control electric field strength: E=E ₀ KV/mm;

(c) Input the acceleration signal into the above-mentioned CDM module again to judge whether the flutter omen still exists at this time, if the flutter omen disappears, make the applied electric field strength E=0, and then the program goes to step (2) to continue running. If the chatter warning still exists, the program runs downward;

(d) Calculate the flutter frequency F _C at this time;

(e) Discriminate FC>F _post ?

If F _C >F _post , set E=E-0.1KV/mm, re-measure the acceleration, and run according to (c)~(d);

If F _C < F _post , set E=E+0.1KV/mm, re-measure the acceleration, and run according to (c)~(d);

(4) The above-mentioned program flow runs continuously during the cutting process until the end of the cutting process, and the control program ends;

The above CDM module performs the following steps in sequence:

(1) Input sampling acceleration signal r _i ;

(2) input r _i into the LO-RBF type neural network structure program module, and run according to the following steps successively:

(a) Calculate the estimated value f(r _i ) of the probability density function of the sensor signal time series at r _i and the estimated value of the first-order differential function of the probability density function at r _i $\frac{\&PartialD;}{\&PartialD;r_i}f\left(r_i\right),$

i.e. calculate:

$<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; \&PartialD;</span>}{<span\; class="notranslate">\; \&PartialD;</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; LO</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; RBF</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$

(b) When constructing a new signal sequence G _n , contrasting with each input sensing signal sampling value r _i , the element g _i in G _n is:

${\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\frac{<span\; class="notranslate">\; \&PartialD;</span>}{<span\; class="notranslate">\; \&PartialD;</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

(3) Input g _i into the Fuzzy ARTMap chatter omen recognition module, and run according to the following steps in turn:

(a) Discriminate i>128? , make a judgment and run it in a loop,

(b) When the sampling frequency is 1000 Hz, when judging i>128, judge whether flutter fmod(i, 50)=0 occurs every 50 milliseconds? Right now

(c) Take out the time signal sequence: take out the first 127 points in the G sequence, and form a time signal sequence G ₁₂₈ of 128 points with the current value g _i , that is, G ₁₂₈ ={g _i-127 , g _i-126 ,...,g _i-1 , g _i },

(d) G ₁₂₈ is input into the fast Fourier transform (FFT) subroutine, obtains the Fourier transform sequence F ₁₂₈ of G ₁₂₈ , F ₁₂₈ =FFT (G ₁₂₈ ),

(e) Taking the modules of the first 64 elements of the F ₁₂₈ sequence to form a new sequence S ₆₄ , each element in S ₆₄ represents the energy density of each frequency point in the analysis frequency band S _i =|F _i |(i=1, 2, 3 ,...64),

(f) 64-dimensional sequence S ₆₄ is input into the Fuzzy ARTMap neuron network (fuzzy adaptive resonance neuron network) subroutine,

(g) Discriminate the two-dimensional vector C ₂ output by the Fuzzy ARTMap neuron network:

C ₂ =Fuzzy ARTMap(S ₆₄ )

If: C ₂ = {0, 1}, it means there is no sign of flutter,

If: C ₂ = {1, 0}, it means there is a sign of flutter,

In the Fuzzy ARTMap supervised learning stage mentioned above, M 64-dimensional signals S ₆₄ and M two-dimensional vectors C ₂ (both {1, 0}) representing chatter signs are input at the same time ART _a sub-module and ART _b sub-module of the network; similarly, the 64-dimensional signal S ₆₄ collected when N flutter signs do not exist and N two-dimensional vectors C ₂ representing no flutter signs (both are {0, 1}) Input the ART _a sub-module and ART _b sub-module of the network respectively at the same time. Here, {0, 1} corresponds to no flutter precursor, and {1, 0} corresponds to flutter precursor. Through such supervised learning, a mapping relationship is established in the form of weights between the ART _a sub-module and the ART _b sub-module. This ensures that the network can judge whether there is a sign of flutter in the acceleration sampling signal based on the input of the ART _a sub-module online.

For the online intelligent control system of machine tool cutting chatter of the present invention, refer to the system block diagram 4 and principle block diagram 5 of the present invention, the system designed by the online intelligent control method of machine tool cutting chatter contains a computer, and is characterized in that it includes an electrorheological material Intelligent boring bar, a voltage converter that applies a high-voltage electric field to the electrorheological material on the boring bar, a computer that applies a control voltage signal to the voltage converter through the motherboard PCI slot and a digital/analog conversion card, and a computer that applies a control voltage signal to the voltage converter through the digital/analog conversion card and The PCI slot of the main board inputs the charge amplifier of the acceleration signal to the computer, and the output terminal is connected with the input terminal of the charge amplifier, and the input signal is the acceleration sensor of the vibration acceleration signal during the cutting process of the end of the boring bar.

In the design of the cutting system, a boring bar containing an electrorheological material is used. When the control electric field intensity applied to the electrorheological material changes from 0 to 2000 volts/mm, the natural frequency of the boring bar has a variation of 30 Hz. .

The electrorheological material configuration process used: mix starch (particle diameter between 10 microns and 50 microns) and vacuum pump oil at a ratio of 9:11, then add an appropriate amount of rosin derivatives and additional additives, and stir with an electric mixer at room temperature for 15 -30 minutes, after stirring, it becomes a uniform brown suspension liquid. After the material is applied with a high-voltage electric field of 0-10000 volts/mm, the shear yield stress it can withstand increases from 0 to 3900 Pa.

The intelligent boring bar containing electrorheological material is on the existing boring bar, with a thin-walled steel ring as the positive electrode of the electrorheological material, and the negative electrode of the electrorheological material is located in the support sleeve relative to the thin-walled steel ring. The electrorheological material is installed in the mm gap, and two O-rings are used to seal the two ends of the electrorheological material. The support sleeve and the boring bar are fixed from two directions perpendicular to each other by four hexagon socket bolts. The support sleeve is clamped in the boring In the tool post in the system, the acceleration sensor is placed at the end of the boring bar.

The present invention introduces smart materials into the on-line control of chatter in metal cutting, and uses the instantaneous response ability of the inherent characteristics of smart materials to electrical control signals, which not only overcomes the shortcoming of the mechanical structure of the cutting system responding slowly to control signals, but also the system obtains On-line suppression of chatter based on the sensing information, which overcomes the difficulty of establishing an accurate control model for the cutting system due to its complexity. Therefore, the invention enhances the ability of the mechanical processing equipment to adapt to different processing objects and processing conditions, and greatly improves the flexibility of the mechanical processing equipment.

And the chatter online control device developed based on this technology can adjust the dynamic characteristics of the processing system in time according to the cutting vibration signal, suppress the cutting chatter in its infancy state, do not produce machining vibration marks on the surface of the workpiece, and do not affect the normal progress of the cutting process , to ensure the quality and efficiency of processing.

In the present invention, the application of CDM online recognition chatter prediction is mainly based on a large number of experimental studies showing that the development process of chatter in the machining process has the following characteristics:

(1) The flutter waveform is similar to the resonant wave, and the growth of the amplitude is a gradual process, which can be divided into the initial flutter stage, the chatter development stage and the full flutter stage.

(2) With the development of chatter, the cutting vibration frequency gradually stabilizes to approach the natural frequency of the system. At this time, the distribution of vibration energy in the frequency domain changes from broadband distribution to narrowband energy distribution.

(3) In the initial flutter stage, the vibration frequency has stabilized to the natural frequency of the system, and the vibration amplitude has not yet reached the maximum amplitude of flutter. It takes approximately 400 to 600 milliseconds or more before the dither amplitude reaches the full dither stage, which gives the monitoring system valuable time for identification and feedback control.

From the above characteristics, it can be known that the resonant wave of flutter and the narrowband characteristics in frequency domain can be used as an important omen of flutter. The flutter prediction technology in the present invention is based on the above analysis, adopts local optimal signal detection technology and neural network technology to identify flutter omens in the flutter development stage, and performs flutter prediction to achieve the purpose of the present invention.

Description of drawings

Fig. 1: Flow chart of the control method in the present invention, F _s = sampling frequency, E = electric field strength, E ₀ = initial electric field strength, AS = acceleration signal, CDM = flutter recognition program module, F _c = flutter frequency, Fpost = flutter frequency when the electrorheological material is in the post-yield state;

Figure 2: Flowchart of CMD’s online identification of flutter signs;

Figure 3: CMD information flow chart;

Fig. 4: system block diagram of the present invention, 1, computer, 2, motherboard PCI slot, 3, analog/digital digital/analog conversion card, 4, analog input port, 5, analog output port, 6, charge amplifier, 7, Acceleration sensor, 8. Voltage converter, 9. Intelligent boring bar (including electrorheological material);

Fig. 5: system block diagram of the present invention, a, interference, b, dynamic cutting force, c, control command, d, vibration response, 10, controller, 11, mechanical vibration system composed of machine tool-workpiece-tool, 12, Cutting process schematic module;

Figure 6: Structural schematic diagram of intelligent boring bar, 13. Support sleeve, 14. Positive electrode, 15. O-ring, 16. Insulation sleeve, 17. Boring bar, 18. Boring tool, L ₁ , length of boring bar clamping, L _2. Overhang length of boring bar;

Figure 7: Schematic diagram of acceleration sensor placement and boring system, 19, chuck, 20, boring tool head fixture, 21, workpiece, 22, boring tool holder;

Figure 8: Schematic diagram of the signal transmission configuration of the monitoring system of the present invention, 23, 8 channel flat adapters;

Figure 9: Schematic diagram of the system device for boring processing using an intelligent boring bar;

Figure 10: Schematic diagram of information input during offline learning and training of Fuzzy-ARTMAP neuron network;

Figure 11: Schematic diagram of information input and output when the Fuzzy-ARTMAP neuron network online predicts flutter omens.

Detailed ways

The present invention can be used in boring machining. Because boring is inner hole processing, the boring bar is generally designed as a slender cantilever beam structure, which has poor rigidity and is prone to bending deformation under force, and chatter is often unavoidable when subjected to dynamic cutting forces. In order to overcome the weakness that the rigidity of the boring bar cannot be fundamentally improved, the present invention adds a smart material——electrorheological material to its structural design. By applying an electric field to the electrorheological material, the overall dynamic characteristics of the boring bar can be changed online, combined with flutter The online prediction technology adjusts the dynamic characteristics of the boring bar online according to the sensor signal to avoid the occurrence of chatter.

In the present embodiment, the system is assembled according to the conventional technology according to Fig. 4, wherein the accessories and interrelationships are described as follows:

(1) The CPU model of the computer is PII233, and the main board has 3 PCI slots, which are used as the carrier of data acquisition and flutter forecast control software in the system.

(2) Analog/digital, digital/analog conversion card: the model is HY-6070, inserted into the PCI slot of the computer, its analog input terminal is connected to the voltage output terminal of the charge amplifier, and the analog output terminal is connected to the input of the high voltage converter connected to the voltage terminal.

(3) Charge amplifier: the model is YE5858. Its input terminal connects the charge output by the acceleration sensor, and the output is a voltage signal representing the acceleration signal. The output is connected to the analog input terminal of the A/D and D/A cards.

(4) Acceleration sensor: The model is YE14103. The acceleration sensor is fixed on the cantilever end of the boring bar through a double-headed screw, and the output is connected to the charge input end of the charge amplifier.

(5) Voltage converter: the model is GYW-010, its function is to convert the low voltage of 0-5 volts output by the computer into a high voltage of 0-10000 volts, and its input voltage terminal is connected with A/D and D/A cards connected to the analog output terminal, and the two poles of the high voltage output terminal are respectively connected to the positive and negative poles of the electrorheological material in the boring bar.

In the structural design of the cutting system, the present invention introduces the electrorheological material into the key components of the mechanical processing system, and develops an intelligent boring bar that can directly control the dynamic characteristics of the mechanical structure by external electrical signals. The structural design of the intelligent boring bar aims at the shortcomings of the limited variation range of the dynamic characteristics of the laminated beam structure or hollow beam structure based on the electrorheological material as the electric field intensity is controlled. The electrorheological material is used to change the local support of the root of the boring bar. The rigidity greatly improves the variation range of the dynamic characteristics of the structure without affecting the normal boring process. The structure of the boring bar is shown in Figure 6. The positive electrode is a thin-walled steel ring, and the part of the support sleeve opposite to the positive electrode is used as the negative electrode of the electrorheological material (that is, the ground). The axial length of the electrode is 20 mm, and the gap between the two electrodes is 0.5 mm. There is an insulating sleeve between the positive electrode and the boring bar, and the electrorheological material is poured into the cavity between the positive electrode and the negative electrode, and the sealing of the electrorheological material is guaranteed by two O-rings. The support sleeve and the boring bar are fixed from two directions perpendicular to each other by four hexagon socket bolts. The support sleeve is installed and clamped in the tool holder in the boring system, L ₁ is equal to 100 mm, L ₂ is equal to 180 mm, and the length to diameter ratio of the overhanging part of the boring bar is 6:1. Install the boring tool on the end of the boring bar. Vibration tests on the boring bar with the vibrator show that the natural frequency of the boring bar has a variation of 30 Hz when the controlled electric field intensity of the electrorheological material varies from 0 to 2000 volts/mm.

The arrangement of boring system and acceleration sensor is shown in Fig. 7. The reason why the acceleration sensor adopts the horizontal direction, that is, the direction perpendicular to the cutting surface is mainly because the cutting chatter is mostly regenerative chatter. From the regenerative chatter generation mechanism, it can be known that only the tool and the workpiece vibrate relative to each other in the direction of cutting depth. It is the displacement component that significantly affects the cutting force. The implementation proves that the vibration response in the direction of cutting depth and the dynamic component of cutting force can reflect the development process of cutting chatter more sensitively than the sensing signals in other directions. Acceleration sensors are installed at the end of the boring bar, because the vibration response is obvious at the end.

The flow chart of the digital signal processed directly by the computer is shown in Figure 8. During processing, the acceleration sensor signal is collected by the YE14103 accelerometer and converted into a voltage signal by the YE5853 charge amplifier, and then input to the 12-bit A/D conversion card through the 8-channel flat cable adapter to become a digital signal that can be directly processed by the computer. .

The configuration of the electrorheological material mainly considers that the electrorheological material works at room temperature, the material system must have high dispersion stability, the material must have a large range of viscosity and elastic modulus, and the electrorheological effect of the material has high stability. To have low conductivity. The basic configuration process is: mix starch and vacuum pump oil, then add appropriate amount of rosin derivatives and additional additives, stir with an electric mixer at room temperature for 10 minutes, and after stirring, it will become a uniform brown suspension liquid. This kind of material has no delamination and precipitation when it is not sealed and placed in the ambient temperature range of 15-40°C, and the electrorheological effect is obvious.

The schematic diagram of the cutting chatter control system is shown in Figure 9. The boring system is built on the CA6140 lathe, the workpiece cantilever is clamped on the spindle, and the boring tool is installed on the tool holder. The monitoring system is a Lenovo PII233 computer equipped with HY6070 data acquisition card. The vibration signal collected is the acceleration signal in the horizontal direction of the end of the boring bar. According to the collected vibration signal and cutting chatter online control strategy, the output channel of the acquisition card sends a control signal to the GYW-010 voltage converter. In this way, the dynamic characteristics of the boring bar can be adjusted by changing the properties of the electrorheological material between the positive and negative electrodes. Such a configuration can ensure that the output voltage of the power supply can be adjusted conveniently according to the cutting vibration signal to control the dynamic characteristics of the boring bar for on-line suppression of cutting chatter.

The operation flow of the system control method is shown in FIG. 1 . The system consists of three parts: system initialization, chatter identification and chatter control.

The system initialization part is mainly to input parameters including sampling frequency and initial electric field intensity determined by experience into the system. The sampling frequency is determined according to the required signal frequency band and the sampling theorem, and the initial electric field strength is selected according to the cutting processing conditions and determined by experiments. Once a flutter precursor is predicted, an initial electric field strength is applied to the electrorheological material. In this way, the time for the system to search for the optimum electric field strength can be saved.

The flutter identification part includes two technical links: preprocessing of vibration signals (such as acceleration, vibration displacement, etc.) and capture of flutter omens in the spectrogram. The preprocessing of the sensing signal is carried out in real time along with the cutting process. The spectrogram of the vibration signal is fed into the Fuzzy ARTMap network every 50 milliseconds for the identification of flutter precursors. During the cutting process, the system continuously monitors the vibration signal, and once the sign of chatter is identified, the online chatter control program will be started.

The third part completes the function of flutter online control, which is divided into two modules. In the first block, first a peak search routine starts to find the chatter frequency F _C in the spectrogram, let F _post equal to F _C . Then, the computer issues a control command to let the control power source of the electro-rheological material apply an initial electric field intensity to the electro-rheological material. Although the initial electric field strength is the optimum value obtained from experiments according to the cutting conditions, it is sometimes necessary to search for the optimum electric field strength online due to the complexity of the cutting system. Therefore, when the flutter omen still exists after the initial electric field intensity is applied, the second module is activated and starts to continuously adjust the electric field intensity on-line until the flutter omen disappears. If the flutter sign disappears, the program jumps to the chatter online identification part, and at the same time, the electric field strength applied to the electrorheological material returns to zero electric field strength.

In the second module, a feedback control strategy is adopted to adjust the electric field strength and suppress the occurrence of flutter. The control strategy is shown in Figure 7: if the flutter omen still exists after the initial electric field strength is applied, the flutter frequency F _C is calculated first, and if the flutter frequency F _C is greater than the flutter frequency F _post before the electric field strength is applied, the flutter frequency F post is reduced The electric field intensity applied to the electro-rheological material is reduced, and if F _C is not greater than F _post , the electric field intensity applied to the electro-rheological material is increased. Then collect the cutting vibration signal to judge whether there is a sign of chatter. If there is a possibility of chatter, then calculate the chatter frequency again, and repeatedly calculate and adjust the electric field strength according to the above process until the sign of chatter disappears.

The information flow of flutter identification and prediction technology is shown in Figure 3. Firstly, the acceleration signal is acquired into the computer through A/D conversion, and the sampling frequency is set to 1000Hz because boring chatter mainly occurs near the natural frequency of the boring bar, generally between 150 and 400Hz, so in order to achieve a satisfactory time-frequency Conversion results, the sampling frequency is taken as 1000Hz.

Then, the acceleration signal sequence is pre-processed by the LO-RBF detection technology to generate a new signal sequence. The purpose of using LO-RBF detection technology is to increase the intensity of the flutter precursor signal in the background noise—the resonance signal. The signal sequence reconstructed by LO-RBF detection technology can take fewer sampling points when doing fast Fourier transform, and its narrow-band characteristics can be displayed in the signal spectrum diagram in the initial stage of flutter. This can reduce a lot of sampling time for data fetching. The present invention confirms through experimentation that adopting 128-point fast Fourier transform can achieve the narrow-band energy distribution characteristic showing the omen of chatter in the acceleration spectrum at the initial stage of boring chatter. Then, the Fuzzy Artmap network model combined with fuzzy theory and adaptive resonance neuron network technology is used as a pattern classifier to judge the stability of the cutting process. During the supervised learning of the network, the input of the network is the modulus vector A (64 dimensions) of each frequency point in the analysis frequency band (1-500Hz) obtained after FFT transformation, and the two-dimensional vector B representing the occurrence of flutter omen and no flutter ({0, 1} and {1, 0}). After successful learning, a definite mapping relationship is established between vector A and vector B. In the online prediction stage, the input of the network is the modulus vector A after the FFT transformation of the real-time online acquisition signal after processing, and the output is the result of forecasting the flutter omen.

In terms of flutter prediction speed, this technology can make accurate prediction within 50ms after the flutter omen appears.

What adopted in the present invention is the Fuzzy ARTMap artificial neuron network model. The Fuzzy-ARTMap network used in the supervised learning stage is composed of a pair of Fuzzy-ART neuron networks (ART _a , ART _{b ,} respectively) and a mapping controller between ART _a , ART _b . Here, the input signal of ART _a is a 64-dimensional signal S ₆₄ , which respectively represent the energy density of each frequency point in the analysis frequency band (1-500 Hz). The input of ART _b is a 2-dimensional signal C ₂ , which is respectively {0, 1} and {1, 0}, which respectively represent flutter omen and no flutter omen.

The learning of the network is performed offline, see Figure 10. Through cutting experiments under different cutting conditions, the sampling sequence of the acceleration signal in the process of flutter from nothing to existence is obtained, and it is reconstructed into a _Gn sequence through the LO-RBF network, and the 128-point FFT transformation is performed on the _Gn sequence. The obtained M 64-dimensional vectors S ₆₄ are used as the input vector group of ART _a . Then according to the depth of the cutting vibration pattern on the surface of the workpiece to judge the moment when the chattering omen appears, the input vector group is divided into two parts: the first K is the input vector group of ART _a when there is no chatter phenomenon during smooth cutting; the latter (MK) is the input vector for ART _a when flutter premonition is present. Corresponding to the above two input vector groups of ART _a , the input vector groups C ₂ of ART _b are {0, 1} and {1, 0} respectively. Fuzzy-ARTMAP adopts fuzzy maximum and minimum learning rules for incremental learning. After successful learning, the weight vectors _{W j} ^{a ,} W k b and W _j connected through the mapping controller between ART _a , ART _b and ART _a , ART _b ^ab establishes a mapping relationship between ART _a and ART _b .

See Figure 11 for the online forecasting stage. At this stage, the input signal of ART _a is the 64-dimensional signal S ₆₄ acquired online in real time, and ART _b is used as the output terminal at this time. When there is no sign of flutter, the output C ₂ is {0, 1}, and when there is a sign of flutter, the output is C ₂ is {1, 0}.

Claims (4)

1, a kind of machine cut flutter on-line intelligence control system, computer is arranged, it is characterized in that: it includes the intelligent boring bar that is provided with er material, er material on boring bar applies the voltage changer of high voltage electric field, apply the computer of controlling voltage signal to voltage changer through mainboard PCI slot and digital-to-analog transition card,, link to each other with the charge amplifier input end to the charge amplifier of computer input acceleration signal through digital-to-analog transition card and mainboard PCI slot with output terminal, and input signal is the acceleration transducer of vibration acceleration signal in the cutting process of boring bar end.

2, machine cut flutter on-line intelligence control system according to claim 1, it is characterized in that, adopted a boring bar that contains a kind of er material in the design of cutting system, when the control electric field strength on being applied to er material is changed to 2000 volts/every millimeter by 0 the boring bar natural frequency has 30 hertz variable quantity.

3, machine cut flutter on-line intelligence control system according to claim 1, it is characterized in that, the er material layoutprocedure that adopts: starch (particle diameter is between 10 microns to 50 microns) and pumping fluid were mixed by 9: 11, add an amount of rosin derivative and additional additives again, at room temperature stirred 15-30 minute with electric blender, stirring the back is uniform brown suspension liquid, this material is after applying 0～10000 volt/every millimeter high voltage electric field, and its anti-shearing yield stress that can bear is increased to 3900 handkerchiefs by 0.

4, machine cut flutter on-line intelligence control system according to claim 1 and 2, it is characterized in that: the described intelligent boring bar that contains er material is on existing boring bar, be with thin-walled steel ring as the er material positive electrode, the er material negative pole is the supporting sleeve that is positioned at respect to thin-walled steel ring part, in the two interelectrode 0.5mm gaps er material is housed, 2 O type circles are adopted in the sealing at er material two ends, supporting sleeve and boring bar are fixed from orthogonal both direction by four hexagon socket head cap screws, supporting sleeve is installed in the knife rest in the boring system, and acceleration transducer is placed in the end of boring bar.

Images