Bearing health assessment method and system based on mamba probability prediction model

Abstract

The application discloses a bearing health assessment method and system based on a Mamba probability prediction model, and belongs to the technical field of bearing component health state assessment. The method comprises the following steps: obtaining an IMF component prediction value by predicting test data from a trained Mamba model, obtaining a prediction frequency sequence by performing fast Fourier transform on the IMF component prediction value, constructing an evaluation model based on the distance between an actual frequency spectrum sequence of train running part bearing vibration data and the prediction frequency sequence, substituting the distance values between each prediction frequency sequence in a train speed sequence to be detected and the actual frequency spectrum sequence of the train running part bearing vibration data into the evaluation model respectively, judging the sigma intervals of all distance values, and obtaining a bearing health state determination result. The application can assess the health states of bearing components such as an outer ring, an inner ring, a roller and a retainer under stable speed, obtain the current health levels of the bearing components, and provide a basis for maintenance.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bearing component health state evaluation, and particularly relates to a bearing health evaluation method and system based on a Mamba probability prediction model. BACKGROUND

[0002] In the current digital era, smart railways have become part of the country's sustainable development. As an important part of smart railways, train intelligent operation and maintenance is an inevitable requirement to improve operational safety, efficiency, service quality and reduce operating costs.

[0003] Among them, the state data such as vibration and temperature of train running gear bearings obtained through the intelligent operation and maintenance system can provide accurate basis for fault early warning and fault alarm. However, it is still difficult to provide a basis for maintenance, that is, it is difficult to evaluate the health status of each component of the bearing. Therefore, a method for evaluating the health status of each component of the bearing, such as the outer ring, inner ring, roller and cage, is needed to obtain the current health level of each component of the bearing and provide a basis for maintenance.

[0004] Currently, the health evaluation of each component of the train running gear bearing is determined by the vibration intensity, and some use LSTM modeling to compare the predicted value with the measured value for evaluation. However, these evaluation methods are only for determining whether there is a fault or not, and the credibility of the evaluation result is not clear. In the monitoring, data processing and prediction process, the LSTM model will forget long-term memory, and the model operation requires a large amount of resources, which is difficult to realize in a vehicle-mounted system. SUMMARY

[0005] The purpose of the present application is to solve the problems in the prior art and provide a bearing health evaluation method and system based on a Mamba probability prediction model to complete the health state evaluation of the bearing components.

[0006] The specific technical solutions adopted by the present application are as follows:

[0007] In order to achieve the above-mentioned application purpose, the present application specifically adopts the following technical solutions:

[0008] In the first aspect, the present application provides a bearing health evaluation system based on a Mamba probability prediction model, which comprises:

[0009] The data monitoring module is used for acquiring train running part bearing vibration data and temperature state data in real time, and comprises a vibration sensor, a temperature sensor, a multiplexing switch, a sensor signal conditioning circuit, a current signal conditioning circuit, a voltage signal conditioning circuit, a voltage division follow-up filter circuit, an A / D converter, a microcontroller module, a wireless transmission module and an overvoltage protection power supply module; the vibration sensor and the temperature sensor are sequentially connected to the microcontroller module through the multiplexing switch, the sensor signal conditioning circuit, the current signal conditioning circuit, the voltage signal conditioning circuit, the voltage division follow-up filter circuit and the A / D converter; and the wireless transmission module and the overvoltage protection power supply module are respectively connected to the microcontroller module.

[0010] The data processing module is used for processing the train running part bearing vibration data and the temperature state data acquired by the data monitoring module, and completing health state evaluation of the bearing part, and comprises a prediction model acquisition module, an evaluation model acquisition module, a health evaluation module and a comprehensive judgment module.

[0011] On the basis of the above scheme, each step can be implemented in the following preferred specific manner.

[0012] As a preferred embodiment of the first aspect, the prediction model acquisition module is configured to use the train running part bearing vibration data and the temperature state data as input data, use a low-pass filter on the input data, and then use a variational mode decomposition to generate n IMF components when there are n spectral frequencies of fault features, use a train speed sequence as input and use each IMF component as output to train a Mamba model;

[0013] The evaluation model acquisition module is configured to obtain a new train speed sequence and form m groups of test data after obtaining the trained Mamba model, input each group of test data into the trained Mamba model for prediction, obtain n IMF component prediction values corresponding to each group of test data, perform a fast Fourier transform on the n IMF component prediction values of each group of test data to obtain a prediction frequency sequence corresponding to each IMF component prediction value, perform a variational mode decomposition and a fast Fourier transform on the train running part bearing vibration data to obtain an actual frequency spectrum sequence corresponding to each IMF component of the train running part bearing vibration data, and sequentially calculate distances between the actual frequency spectrum sequence of the train running part bearing vibration data and each prediction frequency sequence, wherein a distance sequence conforming to a normal distribution is formed by distances corresponding to prediction frequency sequences with the same IMF component prediction value index, an expectation of a normal distribution is estimated as a mean value of each distance sequence, and a variance of a normal distribution is taken as a variance of each distance sequence, and the obtained all normal distributions form an evaluation model.

[0014] The health assessment module is configured to obtain distance values between each of the prediction frequency sequences in the train rotating speed sequence to be detected and the actual frequency spectrum sequence of the train running gear bearing vibration data, and divide each normal distribution into four regions representing bearing health grades according to 1sigma interval, 2sigma interval and 3sigma interval respectively, and obtain the sigma interval where each distance value is located after each distance value is respectively substituted into each normal distribution in the evaluation model; wherein the region corresponding to the 1sigma interval represents that the bearing state is excellent, the region corresponding to the 2sigma interval represents that the bearing state is good, the region corresponding to the 3sigma interval represents that the bearing state is bearing performance reduction, and the other regions represent that the bearing state is bearing failure.

[0015] The comprehensive judgment module is configured to judge the sigma intervals where all the distance values are located: if there is the other region, the bearing failure is taken as the determination result of the bearing health state; if there is no other region and there is the 3sigma interval, the bearing performance reduction is taken as the determination result of the bearing health state; if there is no other region and 3sigma interval and the number of 2sigma intervals exceeds half, the good is taken as the determination result of the bearing health state; if there is no other region and 3sigma interval and the number of 1sigma intervals is greater than or equal to half, the excellent is taken as the determination result of the bearing health state.

[0016] As a preferred embodiment of the first aspect, the microcontroller module comprises a data acquisition control module, a data interpretation module, a power supply circuit, a reset circuit, a crystal oscillator circuit, a data processing and RAM read-write module, an interface chip control unit, a synchronous clock control module, a command frame module, a configuration SPI Flash circuit and a download circuit; wherein the data acquisition control module, the data interpretation module, the power supply circuit, the reset circuit, the crystal oscillator circuit, the interface chip control unit, the synchronous clock control module, the command frame module, the configuration SPI Flash circuit and the download circuit are respectively connected with the data processing and RAM read-write module.

[0017] As a preferred embodiment of the first aspect, the overvoltage protection power module comprises a voltage source and an overvoltage protection circuit connected thereto, the overvoltage protection circuit comprising a first diode, a second diode, a third diode, a fourth diode, a fifth diode, a first resistor, a second resistor, a third resistor, a first PMOS transistor, a second PMOS transistor, a third PMOS transistor and a fourth PMOS transistor; wherein a reference voltage circuit is formed by the first diode, the second diode and the third diode, which is used to generate a reference voltage according to the source voltage when the source voltage exceeds the clamping voltage; a feedback control circuit is formed by the fourth diode, the fifth diode, the first resistor, the second resistor, the third resistor, the first PMOS transistor, the second PMOS transistor, the third PMOS transistor and the fourth PMOS transistor, which is used to receive the reference voltage and clamp the output voltage to the clamping voltage.

[0018] Further, in the overvoltage protection circuit, the anode of the first diode is connected to one end of the first resistor, the cathode of the first diode is connected to the cathode of the second diode, the anode of the second diode is connected to the cathode of the third diode, the anode of the third diode is connected to one end of the second resistor, the anode of the fourth diode, one end of the third resistor, the anode of the fifth diode and the source of the first PMOS transistor respectively, the other end of the first resistor is connected to the drain of the third PMOS transistor, the gate of the third PMOS transistor and the gate of the fourth PMOS transistor respectively, the source of the third PMOS transistor is connected to the source of the fourth PMOS transistor, the drain of the second PMOS transistor and the drain of the first PMOS transistor respectively, the drain of the fourth PMOS transistor is connected to the other end of the second resistor, the cathode of the fourth diode and the gate of the second PMOS transistor respectively, the source of the second PMOS transistor is connected to the other end of the third resistor, the cathode of the fifth diode and the gate of the first PMOS transistor respectively.

[0019] As a preferred embodiment of the first aspect, the sensor signal conditioning circuit comprises an analog signal input terminal, a fourth resistor, a fifth resistor, a first capacitor and a first operational amplifier, one end of the fourth resistor and one end of the fifth resistor are connected to the analog signal input terminal respectively, the other end of the fifth resistor is connected to one end of the first capacitor, the other end of the first capacitor is connected to the other end of the fourth resistor and the positive input terminal of the first operational amplifier respectively, the negative input terminal of the first operational amplifier is connected to the output terminal of the first operational amplifier.

[0020] As a preferred embodiment of the first aspect, the voltage signal conditioning circuit comprises a voltage signal input end, a sixth resistor, a seventh resistor, an eighth resistor, a second capacitor, and a second operational amplifier, one end of the sixth resistor is connected to the voltage signal input end, the other end of the sixth resistor is connected to one end of the seventh resistor and one end of the eighth resistor respectively, one end of the eighth resistor is connected to one end of the second capacitor, the other end of the second capacitor is connected to the other end of the seventh resistor and the positive input end of the second operational amplifier respectively, and the output end of the second operational amplifier is connected to the negative input end of the second operational amplifier.

[0021] As a preferred embodiment of the first aspect, the voltage signal conditioning circuit comprises a voltage signal input end, a sixth resistor, a seventh resistor, an eighth resistor, a second capacitor, and a second operational amplifier, one end of the sixth resistor is connected to the voltage signal input end, the other end of the sixth resistor is connected to one end of the seventh resistor and one end of the eighth resistor respectively, one end of the eighth resistor is connected to one end of the second capacitor, the other end of the second capacitor is connected to the other end of the seventh resistor and the positive input end of the second operational amplifier respectively, and the output end of the second operational amplifier is connected to the negative input end of the second operational amplifier.

[0022] In a second aspect, the application provides a bearing health assessment method based on a Mamba probability prediction model, which comprises the following steps:

[0023] S1. Train speed sequence of the train is taken as input, and the Mamba model is trained by taking the vibration data of the train running part bearing and the temperature state data as input data and using the low-pass filtering and the variational mode decomposition to generate n IMF components, so as to obtain the trained Mamba model;

[0024] S2. After obtaining the trained Mamba model, new train speed sequence is obtained and m groups of test data are constituted, each group of test data is input into the trained Mamba model for prediction, n IMF component prediction values are obtained for each group of test data, each IMF component prediction value corresponds to a prediction frequency sequence after the fast Fourier transform of the n IMF component prediction values of each group of test data, each IMF component of the train running part bearing vibration data corresponds to an actual frequency spectrum sequence after the variational mode decomposition and the fast Fourier transform of the train running part bearing vibration data, the distance between the actual frequency spectrum sequence of the train running part bearing vibration data and each prediction frequency sequence is calculated in turn, the distance corresponding to the prediction frequency sequence with the same IMF component prediction value index constitutes a distance sequence obeying normal distribution, the mean of each distance sequence is estimated as an expectation of a normal distribution, and the variance of each distance sequence is taken as a variance of a normal distribution, and all the normal distributions obtained constitute an evaluation model.

[0025] S3. Obtain the distance value between each predicted frequency sequence in the train rotating speed sequence to be detected and the actual frequency spectrum sequence of the train running gear bearing vibration data, and divide each normal distribution into four regions representing the bearing health level according to the 1sigma interval, 2sigma interval and 3sigma interval respectively, and then obtain the sigma interval where each distance value is located after each distance value is respectively substituted into each normal distribution in the evaluation model; wherein the region corresponding to the 1sigma interval represents that the bearing state is excellent, the region corresponding to the 2sigma interval represents that the bearing state is good, the region corresponding to the 3sigma interval represents that the bearing state is reduced in performance, and the other region represents that the bearing state is bearing failure;

[0026] S4. Judge the sigma interval where all distance values are located: if there is other region, take the bearing failure as the determination result of the bearing health state; if there is no other region and there is 3sigma interval, take the reduced performance as the determination result of the bearing health state; if there is no other region and 3sigma interval, and the number of 2sigma interval exceeds half, take good as the determination result of the bearing health state; if there is no other region and 3sigma interval, and the number of 1sigma interval is greater than or equal to half, take excellent as the determination result of the bearing health state.

[0027] Further, in step S2, the distance dij between the ith actual frequency spectrum sequence of the train running gear bearing vibration data and the predicted frequency sequence corresponding to the ith IMF component prediction value in the jth test data is calculated in the following manner:

[0028] ;

[0029] Wherein, represents the square root; represents the predicted frequency sequence corresponding to the ith IMF component prediction value in the jth test data, respectively represents the ith frequency component in ; represents the ith actual frequency spectrum sequence of the train running gear bearing vibration data, respectively represents the ith frequency component in .

[0030] Compared with the prior art, the present application has the following beneficial effects:

[0031] ​​The application provides a bearing health assessment method and system based on a Mamba probability prediction model, state data such as train running part bearing vibration and temperature of newly collected trains are input into the assessment model, n distances dci between prediction and actual values can be obtained, the i data are matched in the probability density function obtained in step 2, the health state of each component of the bearing, such as the outer ring, the inner ring, the roller and the retainer, is obtained by the method for assessing the health state of each component of the bearing, the health level of each component of the bearing at present is provided, and the basis for maintenance is provided, the assessment is carried out at a stable rotating speed, and the health state assessment of the bearing component is completed. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 It is a step flow chart of the method of the application.

[0033] Figure 2 It is a schematic diagram of the health state assessment of the application.

[0034] Figure 3 It is a structure principle diagram of the data monitoring module of the application.

[0035] Figure 4 It is a structure principle diagram of the microcontroller module of the application.

[0036] Figure 5 It is a circuit diagram of the overvoltage protection power supply module of the application.

[0037] Figure 6 It is a circuit diagram of the sensor signal conditioning circuit of the application.

[0038] Figure 7 It is a circuit diagram of the voltage signal conditioning circuit of the application.

[0039] Figure 8 It is a circuit diagram of the voltage signal conditioning circuit of the application.

[0040] In the figure, first diode Z1, second diode Z2, third diode Z3, fourth diode Z4, fifth diode Z5, first resistor R1, second resistor R2, third resistor R3, fourth resistor R4, fifth resistor R5, sixth resistor R6, seventh resistor R7, eighth resistor R8, ninth resistor R9, tenth resistor R10 and eleventh resistor R11, first capacitor C1, second capacitor C2, third capacitor C3, first PMOS tube M1, second PMOS tube M2, third PMOS tube M3 and fourth PMOS tube M4, first operational amplifier U1, second operational amplifier U2, third operational amplifier U3 and fourth operational amplifier U4. DETAILED DESCRIPTION

[0041] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the drawings. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present application, so the present application is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the present application can be combined accordingly without conflict.

[0042] In the description of the present application, it should be understood that when an element is considered to be "connected" to another element, it can be directly connected to the other element or indirectly connected to the other element with an intermediate element. In contrast, when an element is referred to as being "directly" connected to another element, there is no intermediate element.

[0043] In the description of the present application, it should be understood that the terms "first", "second" are only used for distinguishing purposes and should not be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features.

[0044] In a preferred implementation of the present application, a bearing health assessment system based on a Mamba probability prediction model is provided, which comprises a data monitoring module and a data processing module, the data monitoring module and the data processing module are information interactive, the data processing module includes a prediction model acquisition module, an evaluation model acquisition module, a health assessment module and a comprehensive judgment module.

[0045] As shown in Figure 3 The data monitoring module includes a vibration sensor, a temperature sensor, a multiplexing switch, a sensor signal conditioning circuit, a current signal conditioning circuit, a voltage signal conditioning circuit, a voltage division follow-up filter circuit, an A / D converter, a microcontroller module, a wireless transmission module and an overvoltage protection power supply module. The vibration sensor and the temperature sensor are connected to the microcontroller module in sequence through the multiplexing switch, the sensor signal conditioning circuit, the current signal conditioning circuit, the voltage signal conditioning circuit, the voltage division follow-up filter circuit and the A / D converter. The wireless transmission module and the overvoltage protection power supply module are connected to the microcontroller module respectively.

[0046] It should be noted that in the present application, the vibration sensor and the temperature sensor are used to collect vibration signal parameters and temperature parameters, the chip model of the multiplexing switch is AMC4601, and after the output is selected by the multiplexing analog switch to the signal processing circuit, the current signal conditioning circuit is used for data preprocessing of the signal parameters, the A / D converter is used for digital-to-analog conversion input to the AD7794, and the wireless transmission module is used for data transmission. In the running process of the train, the health status evaluation problem of the bearing parts is effectively monitored, so the present application designs a sensor signal conditioning circuit, a voltage signal conditioning circuit, a voltage division and follow-up filter circuit, an overvoltage protection power supply module and a microcontroller module. The sensor signal conditioning circuit can process the collected sensor signal parameters, the voltage signal conditioning circuit can process the sensor voltage signal, the voltage division and follow-up filter circuit can pass through the voltage division and follow-up filter circuit, and then input into the AD converter. The voltage follower makes the circuit present high resistance input and low resistance output, and improves the load capacity of the circuit. The overvoltage protection power supply module can improve the power supply safety and provide overvoltage protection, and the microcontroller module can control the operation and cooperation of each module in real time. The specific implementation mode of the improved circuit and module is introduced below.

[0047] As shown in Figure 4 The microcontroller module includes a data acquisition control module, a data interpretation module, a power supply circuit, a reset circuit, a crystal oscillator circuit, a data processing and RAM read-write module, an interface chip control unit, a synchronous clock control module, a command frame module, a configuration SPI Flash circuit and a download circuit. The data acquisition control module, the data interpretation module, the power supply circuit, the reset circuit, the crystal oscillator circuit, the interface chip control unit, the synchronous clock control module, the command frame module, the configuration SPI Flash circuit and the download circuit are connected with the data processing and RAM read-write module.

[0048] It should be noted that in the microcontroller module of the application, the data acquisition control module is used for data acquisition control, the data interpretation module is used for data caching, the power supply circuit is used for power supply, the reset circuit is used for reset control, the data processing and RAM read-write module is used for data processing, the interface chip control unit is used for interface control, the synchronous clock control module is used for synchronous clock control, the command frame module is used for command frame control, the configuration SPI Flash circuit is used for serial peripheral interface, and the download circuit is used for data download control. The application adopts the FPGA of the Spartan6 series of Xilinx Company as a core control device, realizes the functions of data acquisition control, data caching, data processing, data storage, data transmission and synchronous clock control, and has the characteristics of high precision, high speed, good reliability, strong real-time performance and low cost; the application is a 16-channel real-time high-speed high-precision synchronous data acquisition system taking the FPGA as a main processor, the sampling frequency in the actual monitoring engineering is 200kHz, the FPGA is used to reasonably control and coordinate the data flow transmission between various modules, and then the functions of real-time, synchronization and high-speed acquisition required by the system are realized.

[0049] As shown in Figure 5 The overvoltage protection power module includes a voltage source and an overvoltage protection circuit connected thereto, and specifically, the overvoltage protection circuit includes a first diode Z1, a second diode Z2, a third diode Z3, a fourth diode Z4, a fifth diode Z5, a first resistor R1, a second resistor R2, a third resistor R3, a first PMOS tube M1, a second PMOS tube M2, a third PMOS tube M3 and a fourth PMOS tube M4. Among them, the reference voltage circuit is composed of the first diode Z1, the second diode Z2 and the third diode Z3, which is used to generate a reference voltage according to the source voltage when the source voltage exceeds the clamping voltage; the feedback control circuit is composed of the fourth diode Z4, the fifth diode Z5, the first resistor R1, the second resistor R2, the third resistor R3, the first PMOS tube M1, the second PMOS tube M2, the third PMOS tube M3 and the fourth PMOS tube M4, which is used to receive the reference voltage and clamp the output voltage to the clamping voltage.

[0050] In the overvoltage protection circuit, an anode of the first diode Z1 is connected to one end of the first resistor R1, a cathode of the first diode Z1 is connected to a cathode of the second diode Z2, an anode of the second diode Z2 is connected to a cathode of the third diode Z3, an anode of the third diode Z3 is connected to one end of the second resistor R2, an anode of the fourth diode Z4, one end of the third resistor R3, an anode of the fifth diode Z5 and a source of the first PMOS M1 respectively, the other end of the first resistor R1 is connected to a drain of the third PMOS M3, a gate of the third PMOS M3 and a gate of the fourth PMOS M4 respectively, a source of the third PMOS M3 is connected to a source of the fourth PMOS M4, a drain of the second PMOS M2 and a drain of the first PMOS M1 respectively, a drain of the fourth PMOS M4 is connected to the other end of the second resistor R2, a cathode of the fourth diode Z4 and a gate of the second PMOS M2 respectively, a source of the second PMOS M2 is connected to the other end of the third resistor R3, a cathode of the fifth diode Z5 and a gate of the first PMOS M1 respectively.

[0051] It should be noted that in the overvoltage protection circuit, the third PMOS M3 and the fourth PMOS M4 form a current mirror for generating a reference current, the second PMOS M2 serves as a driving tube for generating an opening current of the first PMOS M1, and the first PMOS M1 is a discharge tube for releasing a surge current. Figure 5 The voltage value of Vx remains unchanged. Meanwhile, a reference current is generated at the side of the third PMOS M3, and the current is copied to the fourth PMOS M4 through the current mirror, and the current can raise the voltage between the two ends of the second resistor R2, so that the gate of the second PMOS M2 is greater than the threshold voltage, and the second PMOS M2 is opened. After the second PMOS M2 is opened, the driving current generated by the second PMOS M2 passes through the third resistor R3, and the voltage between the two ends of the third resistor R3 is raised, so that the main discharge device (the first PMOS M1) is opened, and the first PMOS M1 can discharge a large amount of current, so that the whole chip is clamped at a certain voltage. Since the size of the first PMOS M1 is large, the second PMOS M2 needs to generate sufficient driving current to open the first PMOS M1.

[0052] Generally, tens of microamperes of current are sufficient to drive the discharge tube (the first PMOS M1). In addition, in order to improve the discharging efficiency of the first PMOS M1, the gate voltage of the first PMOS M1 can be appropriately raised. Raising the instantaneous gate voltage of the first PMOS M1 to more than the maximum gate voltage is beneficial to enhancing the discharging capacity of the first PMOS M1.

[0053] As the source voltage Vsource increases, more current is injected into the second PMOS transistor M2, which further pulls up the gate voltage of the main clamp device (the first PMOS transistor M1), so that more current can be discharged, and in turn the regulator output voltage is adjusted to the clamping voltage.

[0054] It can be seen that through the feedback control mode, the output voltage can be accurately clamped to the desired clamping voltage regardless of the current discharged by the main clamp transistor (the first PMOS transistor M1), and as a result, the final dynamic resistance of the overvoltage protection circuit can be almost zero.

[0055] As Figure 6 shown, the sensor signal conditioning circuit includes an analog signal input end, a fourth resistor R4, a fifth resistor R5, a first capacitor C1, and a first operational amplifier U1, the analog signal input end is connected to one end of the fourth resistor R4 and one end of the fifth resistor R5 respectively, the other end of the fifth resistor R5 is connected to one end of the first capacitor C1, the other end of the first capacitor C1 is connected to the other end of the fourth resistor R4 and the positive input end of the first operational amplifier U1 respectively, and the negative input end of the first operational amplifier U1 is connected to the output end of the first operational amplifier U1.

[0056] It should be noted that in the sensor signal conditioning circuit of the present application, the operational amplifier is a key device of the conditioning circuit, and the ADI company's AD8608 operational amplifier chip is selected in the embodiment, which combines many excellent characteristics, has four-way input and output under single power supply, ensures high speed, low noise and input bias current, and is widely used in various circuits. Since the input impedance of the operational amplifier is generally high, when the input pin is suspended, it is easily disturbed by the external environment, and therefore the fifth resistor R5 can form a loop between the input end and the analog ground when the input pin is suspended, so as to ensure the stability of the operational amplifier.

[0057] As Figure 7 shown, the voltage signal conditioning circuit includes a voltage signal input end, a sixth resistor R6, a seventh resistor R7, an eighth resistor R8, a second capacitor C2, and a second operational amplifier U2, the voltage signal input end is connected to one end of the sixth resistor R6, the other end of the sixth resistor R6 is connected to one end of the seventh resistor R7 and one end of the eighth resistor R8 respectively, the other end of the eighth resistor R8 is connected to one end of the second capacitor C2, the other end of the second capacitor C2 is connected to the other end of the seventh resistor R7 and the positive input end of the second operational amplifier U2 respectively, and the negative input end of the second operational amplifier U2 is connected to the output end of the second operational amplifier U2.

[0058] It should be noted that in the voltage signal conditioning circuit of the application, the rail-to-rail operational amplifier can maximize the input and output voltage swing close to the power supply voltage value, but there is still a large deviation in the case of large current, and since the input voltage range of the chip AD8608 is 0-0.5V, the sixth resistor R6 and the eighth resistor R8 are arranged to form a voltage dividing circuit to reduce the input voltage to below 5V.

[0059] The current collection method is different according to the size of the current and the difference between AC and DC. Common current collection methods include coaxial shunt method, current transformer method, Rogowski coil method, Hall sensor method and sampling resistor method. The Hall sensor method is suitable for AC and DC current measurement and can measure large current. In the application, the current signal conditioning circuit uses ACS714 chip of Allegro Company. The chip is a current sensor with common mode rejection field effect, which is composed of high-precision low-bias linear Hall sensor. The maximum sampling current of ACS714 chip is 5A.

[0060] The FGPA controls the address switching channel of the analog switch to achieve time division multiplexing effect. When switching the channel, the analog switch will affect the change of the capacitive load, and phenomena such as signal oscillation or ringing will occur. The faster the switching speed of the analog switch, the more obvious the phenomenon. Therefore, the selection of the analog switch is particularly important. Through analysis and comparison of various types of analog switches, the ADG706 chip is used in the application.

[0061] As shown in Figure 8 The voltage dividing and following filter circuit includes a third operational amplifier U3, a fourth operational amplifier U4, a third capacitor C3, a ninth resistor R9, a tenth resistor R10 and an eleventh resistor R11. The output end of the third operational amplifier U3 is connected to one end of the tenth resistor R10. The other end of the tenth resistor R10 is connected to one end of the eleventh resistor R11 and the positive input end of the fourth operational amplifier U4, respectively. The other end of the eleventh resistor R11 is connected to an analog ground. The output end of the fourth operational amplifier U4 is connected to one end of the ninth resistor R9. The other end of the ninth resistor R9 is connected to one end of the third capacitor C3 and an A / D converter, respectively. The other end of the third capacitor C3 is connected to a ground.

[0062] It should be noted that in the voltage dividing and following filter circuit of the application, the signal passes through the analog switch, then passes through the voltage dividing and following filter circuit, and then is input into the A / D converter. The voltage dividing and following filter circuit makes the circuit have high resistance input and low resistance output, thereby improving the load capacity of the circuit. The chip model of the third operational amplifier U3 and the fourth operational amplifier U4 is AD8031.

[0063] As shown in Figure 1As shown, in a preferred implementation of the present application, the bearing health assessment method based on the Mamba probability prediction model includes the following S1-S4 steps. The specific implementation process is described below.

[0064] I. Constructing a prediction model

[0065] S1. The train running gear bearing vibration data and temperature state data are used as input data. When the spectral frequency of the fault feature has n, the input data is filtered by low-pass filtering and then VMD (Variational Modal Decomposition) is used to generate n IMF components (i.e. IMF1, IMF2…IMFn). The train speed sequence r(t) is used as input and the Mamba model is trained with each IMF component as output.

[0066] II. Constructing an evaluation model

[0067] S2. After obtaining the trained Mamba model, new train speed sequences are obtained and m groups of test data are constructed. Each group of test data is input into the trained Mamba model for prediction, and n IMF component prediction values are obtained for each group of test data. After FFT (Fast Fourier Transform) is performed on the n IMF component prediction values of each group of test data, each IMF component prediction value corresponds to a prediction frequency sequence, i.e. the n prediction frequency sequences corresponding to the n IMF component prediction values of each group of test data are denoted as Ap1(f), Ap2(f), …, Apn(f). After VMD (Variational Modal Decomposition) and FFT (Fast Fourier Transform) are performed on the train running gear bearing vibration data, each IMF component of the train running gear bearing vibration data corresponds to an actual frequency spectrum sequence, i.e. the train running gear bearing vibration data itself corresponds to n actual frequency spectrum sequences, which are A1(f), A2(f), …, An(f). The distance between the actual frequency spectrum sequence of the train running gear bearing vibration data and each prediction frequency sequence is calculated in turn, the distance corresponding to the prediction frequency sequence with the same IMF component prediction value index constitutes a distance sequence following a normal distribution, and the mean of each distance sequence is estimated as the expectation of a normal distribution, and the variance of each distance sequence is taken as the variance of a normal distribution. All the normal distributions obtained constitute an evaluation model.

[0068] It should be noted that in step S2 of the present application, the distance dij between the i-th actual frequency spectrum sequence of the train running gear bearing vibration data and the prediction frequency sequence corresponding to the i-th IMF component prediction value in the j-th group of test data is calculated as follows:

[0069] ;

[0070] wherein, denotes the square root; denotes the prediction frequency sequence corresponding to the i-th IMF component prediction value in the j-th group of test data, denote the first frequency components in the denote the i-th actual frequency spectrum sequence of the train running gear bearing vibration data, denote the first frequency components in the

[0071] III. Health assessment

[0072] S3. Obtain the distance values between each prediction frequency sequence in the train speed sequence to be detected and the actual frequency spectrum sequence of the train running gear bearing vibration data, and according to the 1sigma interval, 2sigma interval, and 3sigma interval, respectively divide each normal distribution into four regions representing the bearing health level, and after each distance value is respectively substituted into each normal distribution in the evaluation model, obtain the sigma interval where each distance value is located; wherein the region corresponding to the 1sigma interval (i.e. ① in the above formula) represents that the bearing state is excellent, the region corresponding to the 2sigma interval (i.e. ② in the above formula) represents that the bearing state is good, the region corresponding to the 3sigma interval (i.e. ③ in the above formula) represents that the bearing state is reduced bearing performance, and the other region (i.e. ④ in the above formula) represents that the bearing state is bearing failure. Figure 2 Figure 2 Figure 2 Figure 2

[0073] IV. Comprehensive judgment

[0074] S4. After step S3, judge the sigma intervals where all distance values are located: if there is an other region, take bearing failure as the determination result of the bearing health state; if there is no other region and there is a 3sigma interval, take reduced bearing performance as the determination result of the bearing health state; if there is no other region and 3sigma interval, and the number of 2sigma intervals exceeds half, take good as the determination result of the bearing health state; if there is no other region and 3sigma interval, and the number of 1sigma intervals is greater than or equal to half, take excellent as the determination result of the bearing health state.

[0075] The above-described embodiments are only a preferred scheme of the present application, and are not intended to limit the present application. Those skilled in the related art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, any technical scheme obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present application.​​​​​​

Claims

1. A bearing health assessment system based on the Mamba probabilistic prediction model, characterized in that, include: The data monitoring module is used to acquire real-time vibration data and temperature status data of the train running gear bearings. It includes a vibration sensor, a temperature sensor, a multiplexer, a sensor signal conditioning circuit, a current signal conditioning circuit, a voltage signal conditioning circuit, a voltage divider follower filter circuit, an A / D converter, a microcontroller module, a wireless transmission module, and an overvoltage protection power supply module. The vibration sensor and temperature sensor are connected to the microcontroller module after passing through the multiplexer, sensor signal conditioning circuit, current signal conditioning circuit, voltage signal conditioning circuit, voltage divider follower filter circuit, and A / D converter in sequence. The wireless transmission module and the overvoltage protection power supply module are respectively connected to the microcontroller module. The data processing module is used to process the vibration data and temperature status data of the train running gear bearings obtained by the data monitoring module, and to complete the health status assessment of the bearing components. It includes a prediction model acquisition module, an assessment model acquisition module, a health assessment module, and a comprehensive judgment module. The prediction model acquisition module is used to take the vibration data of the train running gear bearing and the temperature status data as input data. When there are n spectral frequencies of the fault features, the input data is low-pass filtered and then variational mode decomposition is used to generate n IMF components. The train speed sequence is used as input and each IMF component is used as output to train the Mamba model. The evaluation model acquisition module is used to acquire new train speed sequences and form m sets of test data after obtaining the trained Mamba model. Each set of test data is input into the trained Mamba model for prediction. Each set of test data corresponds to n IMF component prediction values. After performing fast Fourier transform on the n IMF component prediction values ​​of each set of test data, each IMF component prediction value corresponds to a prediction frequency sequence. After performing variational mode decomposition and fast Fourier transform on the vibration data of the train running gear bearing, each IMF component of the vibration data of the train running gear bearing corresponds to an actual spectrum sequence. The distance between the actual spectrum sequence of the vibration data of the train running gear bearing and each prediction frequency sequence is calculated in sequence. The distances corresponding to the prediction frequency sequences with the same IMF component prediction value index constitute a distance sequence that follows a normal distribution. The mean of each distance sequence is estimated as the expectation of a normal distribution, and the variance of each distance sequence is taken as the variance of a normal distribution. All the obtained normal distributions constitute the evaluation model. The health assessment module is used to obtain the distance value between each predicted frequency sequence in the train speed sequence to be detected and the actual spectrum sequence of the train running gear bearing vibration data. After substituting each distance value into each normal distribution in the assessment model, the sigma interval of each distance value is obtained. The comprehensive judgment module is used to judge the sigma interval in which all distance values ​​are located, and form a judgment result of the bearing health status.

2. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 1, characterized in that, The health assessment module divides each normal distribution into four regions representing the bearing health level according to the 1 sigma interval, 2 sigma interval, and 3 sigma interval. The region corresponding to the 1 sigma interval represents the bearing condition as excellent, the region corresponding to the 2 sigma interval represents the bearing condition as good, the region corresponding to the 3 sigma interval represents the bearing condition as reduced bearing performance, and the other regions represent the bearing condition as bearing failure. When the comprehensive judgment module judges the sigma intervals of all distance values: if other regions exist, the bearing failure is judged as the bearing health status; if no other regions exist and a 3sigma interval exists, the bearing performance degradation is judged as the bearing health status; if no other regions and a 3sigma interval exist simultaneously, and the number of 2sigma intervals exceeds half, the good bearing health status is judged as the bearing health status; if no other regions and a 3sigma interval exist simultaneously, and the number of 1sigma intervals is greater than or equal to half, the excellent bearing health status is judged as the bearing health status.

3. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 1, characterized in that, The microcontroller module includes a data acquisition and control module, a data interpretation module, a power supply circuit, a reset circuit, a crystal oscillator circuit, a data processing and RAM read / write module, an interface chip control unit, a synchronous clock control module, a command frame decomposition module, a configuration SPI Flash circuit, and a download circuit; wherein the data acquisition and control module, the data interpretation module, the power supply circuit, the reset circuit, the crystal oscillator circuit, the interface chip control unit, the synchronous clock control module, the command frame decomposition module, the configuration SPI Flash circuit, and the download circuit are each connected to the data processing and RAM read / write module.

4. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 1, characterized in that, The overvoltage protection power supply module includes a voltage source and an overvoltage protection circuit connected thereto. The overvoltage protection circuit includes a first diode (Z1), a second diode (Z2), a third diode (Z3), a fourth diode (Z4), a fifth diode (Z5), a first resistor (R1), a second resistor (R2), a third resistor (R3), a first PMOS transistor (M1), a second PMOS transistor (M2), a third PMOS transistor (M3), and a fourth PMOS transistor (M4). The first diode (Z1), the second diode (Z2), and the third diode (Z3) form a reference voltage circuit, which generates a reference voltage based on the source voltage when the source voltage exceeds the clamping voltage. The fourth diode (Z4), the fifth diode (Z5), the first resistor (R1), the second resistor (R2), the third resistor (R3), the first PMOS transistor (M1), the second PMOS transistor (M2), the third PMOS transistor (M3), and the fourth PMOS transistor (M4) form a feedback control circuit, which receives the reference voltage and clamps the output voltage to the clamping voltage.

5. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 4, characterized in that, In the overvoltage protection circuit, the anode of the first diode (Z1) is connected to one end of the first resistor (R1), the cathode of the first diode (Z1) is connected to the cathode of the second diode (Z2), the anode of the second diode (Z2) is connected to the cathode of the third diode (Z3), the anode of the third diode (Z3) is connected to one end of the second resistor (R2), the anode of the fourth diode (Z4), one end of the third resistor (R3), the anode of the fifth diode (Z5), and the source of the first PMOS transistor (M1), respectively. The other end of the first resistor (R1) is connected to the drain of the third PMOS transistor (M3) and the source of the third PMOS transistor (M1). The gate of the S-channel transistor (M3) and the gate of the fourth PMOS transistor (M4) are connected. The source of the third PMOS transistor (M3) is connected to the source of the fourth PMOS transistor (M4), the drain of the second PMOS transistor (M2), and the drain of the first PMOS transistor (M1). The drain of the fourth PMOS transistor (M4) is connected to the other end of the second resistor (R2), the cathode of the fourth diode (Z4), and the gate of the second PMOS transistor (M2). The source of the second PMOS transistor (M2) is connected to the other end of the third resistor (R3), the cathode of the fifth diode (Z5), and the gate of the first PMOS transistor (M1).

6. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 1, characterized in that, The sensor signal conditioning circuit includes an analog signal input terminal, a fourth resistor (R4), a fifth resistor (R5), a first capacitor (C1), and a first operational amplifier (U1). The analog signal input terminal is connected to one end of the fourth resistor (R4) and one end of the fifth resistor (R5). The other end of the fifth resistor (R5) is connected to one end of the first capacitor (C1). The other end of the first capacitor (C1) is connected to the other end of the fourth resistor (R4) and the positive input terminal of the first operational amplifier (U1). The negative input terminal of the first operational amplifier (U1) is connected to the output terminal of the first operational amplifier (U1).

7. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 1, characterized in that, The voltage signal conditioning circuit includes a voltage signal input terminal, a sixth resistor (R6), a seventh resistor (R7), an eighth resistor (R8), a second capacitor (C2), and a second operational amplifier (U2). The voltage signal input terminal is connected to one end of the sixth resistor (R6). The other end of the sixth resistor (R6) is connected to one end of the seventh resistor (R7) and one end of the eighth resistor (R8). The other end of the eighth resistor (R8) is connected to one end of the second capacitor (C2). The other end of the second capacitor (C2) is connected to the other end of the seventh resistor (R7) and the positive input terminal of the second operational amplifier (U2). The negative input terminal of the second operational amplifier (U2) is connected to the output terminal of the second operational amplifier (U2).

8. The bearing health assessment system based on the Mamba probabilistic prediction model as described in claim 1, characterized in that, The voltage divider follower filter circuit includes a third operational amplifier (U3), a fourth operational amplifier (U4), a third capacitor (C3), a ninth resistor (R9), a tenth resistor (R10), and an eleventh resistor (R11). The output terminal of the third operational amplifier (U3) is connected to one end of the tenth resistor (R10). The other end of the tenth resistor (R10) is connected to one end of the eleventh resistor (R11) and the positive input terminal of the fourth operational amplifier (U4). The other end of the eleventh resistor (R11) is connected to analog ground. The output terminal of the fourth operational amplifier (U4) is connected to one end of the ninth resistor (R9). The other end of the ninth resistor (R9) is connected to one end of the third capacitor (C3) and the A / D converter. The other end of the third capacitor (C3) is grounded.

9. A bearing health assessment method based on the Mamba probabilistic prediction model, characterized in that, Includes the following steps: S1. Take the vibration data and temperature status data of the train running gear bearing as input data. When there are n spectral frequencies of the fault characteristics, the input data is low-pass filtered and then variational mode decomposition is used to generate n IMF components. The train speed sequence is taken as input and each IMF component is taken as output to train the Mamba model and obtain the trained Mamba model. S2. After obtaining the trained Mamba model, acquire new train speed sequences and form m sets of test data. Input each set of test data into the trained Mamba model for prediction. Each set of test data corresponds to n IMF component prediction values. After performing Fast Fourier Transform on the n IMF component prediction values ​​of each set of test data, each IMF component prediction value corresponds to a prediction frequency sequence. After performing Variational Mode Decomposition and Fast Fourier Transform on the vibration data of the train running gear bearing, each IMF component of the vibration data of the train running gear bearing corresponds to an actual spectrum sequence. Calculate the distance between the actual spectrum sequence of the vibration data of the train running gear bearing and each prediction frequency sequence in turn. The distances corresponding to the prediction frequency sequences with the same IMF component prediction value index constitute a distance sequence that follows a normal distribution. Estimate the mean of each distance sequence as the expectation of a normal distribution and take the variance of each distance sequence as the variance of a normal distribution. All the obtained normal distributions constitute the evaluation model. S3. Obtain the distance values ​​between each predicted frequency sequence in the train speed sequence to be detected and the actual spectrum sequence of the train running gear bearing vibration data. Divide each normal distribution into four regions representing the bearing health level according to the 1 sigma interval, 2 sigma interval, and 3 sigma interval. Substitute each distance value into each normal distribution in the evaluation model to obtain the sigma interval in which each distance value is located. Among them, the region corresponding to the 1 sigma interval represents the bearing condition as excellent, the region corresponding to the 2 sigma interval represents the bearing condition as good, the region corresponding to the 3 sigma interval represents the bearing condition as reduced bearing performance, and the other regions represent the bearing condition as bearing failure. S4. Determine the sigma interval for all distance values: If other regions exist, the bearing failure is determined as the bearing health status; if no other regions exist and a 3sigma interval exists, the bearing performance degradation is determined as the bearing health status; if no other regions and 3sigma intervals exist simultaneously, and the number of 2sigma intervals exceeds half, the good bearing health status is determined as the bearing health status; if no other regions and 3sigma intervals exist simultaneously, and the number of 1sigma intervals is greater than or equal to half, the excellent bearing health status is determined as the bearing health status.

10. The bearing health assessment method based on the Mamba probabilistic prediction model as described in claim 9, characterized in that, In step S2, the distance dij between the i-th actual spectrum sequence of the train running gear bearing vibration data and the predicted frequency sequence corresponding to the i-th IMF component prediction value in the j-th test data is calculated as follows: ； in, To represent the square root; This represents the predicted frequency sequence corresponding to the predicted value of the i-th IMF component in the j-th test data set. They represent The first in One frequency component; This represents the i-th actual spectral sequence of vibration data from the bearings of the train's running gear. They represent The first in Each frequency component.

Images