Direct current steering motor encoder disconnection detection method

Abstract

The invention discloses a direct current steering motor encoder broken line detection method, relates to the technical field of encoder broken line detection, and solves the problems that in the prior art, the encoder speed judgment instruction is large, the feedback speed is lower than a certain rotating speed value, and errors are reported after long-time duration. The encoder broken line detection method comprises the following steps: simulating loop characteristic data of a steering motor under all working conditions; establishing a deep belief network model of the steering motor according to the motor simulation data feature library; based on the deep belief network model and in combination with the loop feature data, generating a dynamic discrimination threshold of the encoder disconnection fault; according to the operation data flow of the steering motor and in combination with the dynamic discrimination threshold, identifying an encoder disconnection fault of the steering motor; performing disturbance compensation on the loop data stream by the environment data stream of the steering motor, and identifying a transient condition in combination with frequency domain characteristics; and after the encoder has a broken line fault, the controller starts a self-healing execution mechanism to regulate and control the steering motor, the deep belief network model is updated and optimized, and encoder broken line detection is completed.

Description

Technical Field

[0001] The present invention relates to the technical field of encoder wire break detection, and particularly to a method for detecting wire breaks in an encoder of a DC steering motor. Background Art

[0002] With the comprehensive transformation of the automotive industry towards Electric Power Steering (EPS), as the core actuator, the reliable operation of the DC steering motor is directly related to vehicle handling safety. In EPS, the encoder plays a key role in real-time feedback of the motor rotor position and speed, and its signal accuracy directly affects the stability of closed-loop control. However, under complex working conditions (such as extreme temperatures, mechanical vibrations, or wire harness aging), encoder wire break faults may lead to signal loss or distortion, thereby causing serious consequences such as steering assist failure, steering wheel jitter, or even vehicle out of control. According to statistics, approximately 15% of vehicle electronic system failures are related to abnormal sensor signals, and the problem of encoder wire breaks is particularly prominent in frequently used steering systems.

[0003] Traditional detection methods mainly rely on hardware redundancy design (such as dual encoder configuration) or periodic self-check programs. However, in cost-sensitive mass-produced vehicle models, adding redundant hardware significantly increases the material cost and installation complexity. At the same time, software detection algorithms based on fixed thresholds are prone to false alarms and missed detections under dynamic load changes (such as torque differences between parking steering and high-speed lane changes) or transient interferences (electromagnetic noise), and it is difficult to meet the requirements for fault diagnosis coverage in the ISO 26262 functional safety standard.

[0004] Patent No. CN202410050243.X discloses an absolute encoder wire break detection method, device, electronic device, and storage medium, which relates to the technical field of encoder detection. The method includes: obtaining motor parameters characterizing the motor operating state; determining a first initial position of the rotor of the motor through a preset pulsating square wave high-frequency injection strategy according to the motor parameters; inputting the motor parameters into a preset electrical angle observer to determine a first operating position of the rotor; and determining whether the absolute encoder is broken according to the first initial position and the first operating position. In this way, the problems of poor versatility, high manufacturing cost, high failure rate, and low reliability existing in traditional absolute encoder wire break detection can be improved.

[0005] Patent No. CN202080107565.3 discloses a power conversion device, a motor drive system, and a method for detecting a break in a signal transmission cable. The power conversion device (10) includes: a power conversion circuit (14) that outputs the power input to the power input terminals (11U, 11V, 11W, 11P, 11N) to the power output terminals (12U, 12V, 12W) connected to the power cables (21U, 21V, 21W) of the power cable (20) for supplying alternating current to the motor (60); a ground terminal (12G) that is connected to the ground wire (21G) of the power cable (20); an encoder circuit (15) that outputs encoder information based on the angle information of the signal lines (31a, 31b) of the signal transmission cable (30) input to the information input terminals (13a, 13b); a control unit (16) that outputs a control signal for controlling the power conversion circuit (14) based on the encoder information from the encoder circuit (15); a shield terminal (13G) that is connected to the shield (31G) of the signal transmission cable (30) electrically connected to the ground wire (21G) of the power cable (20) on the motor (60) side; and a determination unit (17) that detects the current flowing through the shield (30G) of the signal transmission cable (30) input to the shield terminal (13G) and outputs break detection information for a break in the shield (30G) of the signal transmission cable (30).

[0006] Although the above patent can detect a break in the encoder of the DC steering motor of a vehicle; in the prior art, it is judged that the encoder speed command is large, and the feedback speed is always lower than a certain rotational speed value and an error is reported after a long time. Due to the long time, the steering rotation angle will be too large. If the judgment time is only reduced, false alarms are likely to occur. Summary of the Invention

[0007] The object of the present invention is to provide a method for detecting a break in the encoder of a DC steering motor, which can generate multi-level decisions for warning of the encoder break fault of the steering motor through a deep belief network model combined with a dynamic discrimination threshold; so as to improve the timeliness and safety of fault notification when the encoder breaks.

[0008] The present invention utilizes the following technical solutions: A method for detecting a break in the encoder of a DC steering motor includes the following steps; S1: Simulate the loop characteristic data of the steering motor under all working conditions and establish a motor simulation data feature library; All working conditions include normal working conditions and abnormal working conditions; S2: Establish a deep belief network model of the steering motor according to the motor simulation data feature library; S3: Based on the deep belief network model and combined with the loop characteristic data, generate a dynamic discrimination threshold for encoder disconnection faults; S4: Based on the operation data stream of the steering motor and combined with the dynamic discrimination threshold, identify the encoder disconnection fault of the steering motor; The operation data stream includes the environmental data stream and the loop data stream; S5: Combine the environmental data stream of the steering motor to perform disturbance compensation on the loop data stream, and identify the transient working condition by combining the frequency domain characteristics; S6: After the encoder disconnection fault occurs, the controller starts a self-healing execution mechanism to regulate the steering motor, and then updates and optimizes the deep belief network model to complete the encoder disconnection detection.

[0009] Preferably, step S1 includes the following steps: S11: Extract the loop commands of the steering motor from the controller; the loop commands include the preset position, preset speed, and preset current amplitude; the loop includes a position loop, a speed loop, and a current loop; S12: Simulate the loop characteristic data of the steering motor under normal working conditions; the loop characteristic data includes the simulated position, simulated speed, and simulated current amplitude; S13: Measure the simulated difference between the loop characteristic data and the corresponding loop commands; the simulated difference includes the position deviation, speed frequency error, and fluctuation amplitude error; S14: According to the simulated difference and the corresponding judgment threshold, obtain the loop characteristic type of the steering motor under abnormal working conditions.

[0010] Preferably, step S1 also includes the following steps: If the position deviation is greater than the position threshold and the duration reaches the preset value, the loop characteristic type of the abnormal working condition of the position loop is position integral saturation; if the position deviation is less than or equal to the position threshold but the duration does not reach the preset value, the position loop is in normal working condition; If the speed frequency error is greater than the frequency threshold, the loop characteristic type of the abnormal working condition of the speed loop is speed frequency oscillation; if the speed frequency error is less than or equal to the frequency threshold, the speed loop is in normal working condition; If the fluctuation amplitude error is greater than the fluctuation threshold, the loop characteristic type of the abnormal working condition of the current loop is current overmodulation; if the fluctuation amplitude error is less than or equal to the fluctuation threshold, the current loop is in normal working condition; Establish a motor simulation data feature library according to the loop characteristic data of the steering motor under normal working conditions and the loop characteristic types under abnormal working conditions.

[0011] Preferably, the preset sub-packaging mechanism continuously performs real-time sub-packaging on the upgrade package during the transmission of the package to be deployed according to the real-time network signal strength, and uses the hash algorithm to generate independent check codes based on the sub-packaging time, sub-packaging quantity, and overall cyclic redundancy check code.

[0012] Preferably, step S2 includes the following steps: S21: Establish a joint domain measurement layer to perform time-domain analysis and frequency-domain analysis on the loop characteristic data respectively, and obtain the mean value, standard deviation, variance, peak-to-peak value and spectrum energy distribution map of each loop; S22: Establish a feature decomposition layer to perform wavelet packet decomposition on the loop characteristic data, and obtain the data energy entropy of each loop; S23: Establish a feature extraction layer to extract features from the loop characteristic data and the loop characteristic type, and obtain the original feature matrix; S24: Establish a multi-dimensional training layer, and perform several cross-scale trainings on the original feature matrix in combination with time domain, frequency domain and wavelet packet decomposition to obtain a time-frequency domain joint weight matrix: S25: According to the simulation difference and the judgment threshold, the identification update layer performs optimization feedback on the time-frequency domain joint weight matrix, and then establishes a deep belief network model for the steering motor.

[0013] Preferably, step S3 includes the following steps: S31: The deep belief network model trains and corrects the loop characteristic data to obtain a covariance matrix; S32: Calculate the feature deviation degree of each loop according to the covariance matrix combined with the mean value of each loop; S33: Obtain the initial judgment threshold according to the standard deviation and mean value of each loop combined with the normal distribution criterion; S34: According to the sliding window adaptive mechanism and the time-frequency domain joint weight matrix, combine the forgetting factor to update the initial judgment threshold in real time to obtain a dynamic discrimination threshold for encoder disconnection faults; S35: Calculate the prediction confidence of the deep belief network model according to the dynamic discrimination threshold.

[0014] Preferably, step S4 includes the following steps: S41: Use a sliding synchronous window to synchronize the position, speed, current, voltage and temperature of the steering motor to obtain an operation data stream; S42: Clean, associate and classify the operation data stream to obtain a loop data stream and an environment data stream; S43: Input the loop data stream into the deep belief network model, and combine the dynamic discrimination threshold to give an early warning of the encoder disconnection fault of the steering motor and generate multi-level decisions.

[0015] Preferably, step S43 includes the following steps: S431: If all loop characteristics in the loop data stream are less than the corresponding dynamic discrimination threshold, it is determined that the encoder is operating normally, and a green early warning is displayed; S432: If any loop feature in the loop data stream is greater than or equal to the corresponding dynamic discrimination threshold and lasts for N cycles, send a yellow warning of encoder disconnection fault to the administrator; S433: If any two loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence is greater than or equal to the sub-layer determination threshold, send an orange warning of encoder disconnection fault to the administrator; S434: If all loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence is greater than or equal to the first-layer determination threshold, send a red warning of encoder disconnection fault to the administrator.

[0016] Preferably, step S5 includes the following steps: S51: Perform a fitting operation on the temperature and voltage of the environmental data stream to obtain a temperature curve, a voltage curve, and a temperature-voltage fitting formula; S52: Perform a disturbance compensation operation on the loop data stream according to the temperature curve in time sequence to obtain a loop temperature-compensated data stream; S53: Perform a fluctuation correction operation on the loop temperature-compensated data stream according to the voltage curve in frequency to obtain a loop comprehensive data stream; S54: Verify the loop comprehensive data stream according to the temperature-voltage fitting formula and convert it to the frequency domain to obtain the frequency domain characteristics of each loop, and judge the transient working condition of the steering motor.

[0017] Preferably, step S5 further includes the following steps: If the energy in the low-frequency band of the frequency domain characteristics is concentrated, judge the transient working condition as acceleration or deceleration according to the current error sign; If the frequency domain characteristics present characteristic harmonics, judge the transient working condition as straightening or lane change according to the current error sign; If the frequency domain characteristics present broadband shock, judge the transient working condition as load mutation; If the frequency domain characteristics present narrowband energy concentration, judge the transient working condition as high-frequency increase or decrease according to the current error sign; If the frequency domain characteristics present an inherent frequency spike of periodic fluctuation, judge the transient working condition as mechanical resonance.

[0018] The present invention generates multi-level decisions on the encoder disconnection fault of the steering motor through a deep belief network model combined with a dynamic discrimination threshold; by calculating the simulated difference between the loop feature data and the corresponding loop instruction, and comparing it with the corresponding judgment threshold, the loop feature type of the steering motor under abnormal working conditions is obtained; the timeliness and safety of fault notification when the encoder is disconnected are improved. Description of the Drawings

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the related art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the related art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained according to the provided accompanying drawings.

[0020] Figure 1 It is a principle block diagram of a method for detecting encoder disconnection of a steering motor; Figure 2 It is a control strategy block diagram of a method for detecting encoder disconnection of a steering motor; Figure 3 It is a control flow chart of a method for detecting encoder disconnection of a steering motor. Detailed implementation manners

[0021] The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments: As Figures 1 - 3 shown, the method for detecting encoder disconnection of the DC steering motor described in the present invention includes the following steps: S1: Simulate the loop characteristic data of the steering motor under all working conditions, and establish a motor simulation data characteristic library; In this embodiment, all working conditions include normal working conditions and abnormal working conditions; S2: Establish a deep belief network model of the steering motor according to the motor simulation data characteristic library; S3: Generate a dynamic discrimination threshold for encoder disconnection faults based on the deep belief network model combined with the loop characteristic data; S4: Identify the encoder disconnection fault of the steering motor according to the operation data stream of the steering motor in combination with the dynamic discrimination threshold; In this embodiment, the operation data stream includes an environmental data stream and a loop data stream; S5: Perform perturbation compensation on the loop data stream in combination with the environmental data stream of the steering motor, and identify transient working conditions in combination with frequency domain characteristics; S6: After the encoder disconnection fault occurs, the controller starts a self-healing execution mechanism to regulate the steering motor, and then updates and optimizes the deep belief network model to complete the encoder disconnection detection.

[0022] In this embodiment, the self-healing execution mechanism is started after the fault is triggered: Immediately switch to the open-loop speed control mode; Start the backup sensor fusion algorithm (Hall signal + back electromotive force estimation); Record the data 10 s before the fault for subsequent analysis.

[0023] In the present invention, step S1 includes the following steps: S11: Extract the loop command of the steering motor from within the controller; The loop command includes a preset position, a preset rotational speed, and a preset current amplitude; the loop includes a position loop, a speed loop, and a current loop; S12: Simulate the loop characteristic data of the steering motor under normal operating conditions; the loop characteristic data includes an analog position, an analog rotational speed, and an analog current amplitude; S13: Measure the analog difference between the loop characteristic data and the corresponding loop command; the analog difference includes a position deviation, a rotational speed frequency error, and a fluctuation amplitude error; S14: Based on the analog difference and the corresponding judgment threshold, obtain the loop characteristic type of the steering motor under abnormal operating conditions.

[0024] In the present invention, the following steps are further included in step S1: If the position deviation is greater than the position threshold and the duration reaches the preset value, the loop characteristic type of the abnormal operating condition of the position loop is position integral saturation; if the position deviation is less than or equal to the position threshold but the duration does not reach the preset value, the position loop is in a normal operating condition; If the rotational speed frequency error is greater than the frequency threshold, the loop characteristic type of the abnormal operating condition of the speed loop is rotational speed frequency oscillation; if the rotational speed frequency error is less than or equal to the frequency threshold, the speed loop is in a normal operating condition; If the fluctuation amplitude error is greater than the fluctuation threshold, the loop characteristic type of the abnormal operating condition of the current loop is current overmodulation; if the fluctuation amplitude error is less than or equal to the fluctuation threshold, the current loop is in a normal operating condition; Based on the loop characteristic data of the steering motor under normal operating conditions and the loop characteristic type under abnormal operating conditions, establish a motor simulation data feature library.

[0025] In this embodiment, position loop integral saturation phenomenon (deviation value > 5° for 200 ms); speed loop oscillation characteristic (abnormal harmonic with frequency > 500 Hz); current loop overmodulation phenomenon (instantaneous value of Iq exceeds the rated value by 150%); In the present invention, the following steps are included in step S2: S21: Establish a joint domain measurement layer to perform time domain analysis and frequency domain analysis on the loop characteristic data respectively, and obtain the mean value, standard deviation, variance, peak-to-peak value, and spectral energy distribution diagram of each loop; In this embodiment, the joint domain measurement layer includes a 6-dimensional time domain feature vector, a 12-dimensional frequency domain feature vector, and a 24-dimensional time-frequency cross feature vector; S22: Establish a feature decomposition layer to perform wavelet packet decomposition on the loop characteristic data, and obtain the data energy entropy of each loop; In this embodiment, the feature decomposition layer includes a 6-dimensional wavelet packet feature vector and an 18-dimensional time-frequency cross feature vector; S23: Establish a feature extraction layer to extract features from the loop feature data and loop feature types, obtaining an original feature matrix; In this embodiment, the feature extraction layer includes 4 convolutional layers with a stride of 2 , 3 convolutional layers with a stride of 1 , and 3 average pooling layers with a stride of 2; S24: Establish a multi-dimensional training layer to perform several cross-scale trainings on the original feature matrix by combining the time domain, frequency domain, and wavelet packet decomposition, obtaining a time-frequency domain joint weight matrix: In this embodiment, the multi-dimensional training layer includes 3 layers of restricted Boltzmann machines; S25: According to the simulation difference and the judgment threshold, the recognition and update layer performs an optimization feedback on the time-frequency domain joint weight matrix, and then establishes a deep belief network model for the steering motor.

[0026] In this embodiment, the recognition and update layer includes a dropout layer, a fully connected layer, and a cross-entropy loss function; In the present invention, step S3 includes the following steps: S31: The deep belief network model performs training and correction on the loop feature data to obtain a covariance matrix; S32: According to the covariance matrix and the mean value of each loop, calculate the feature deviation degree of each loop; S33: According to the standard deviation and mean value of each loop and the normal distribution criterion, obtain an initial judgment threshold; S34: According to the sliding window adaptive mechanism and the time-frequency domain joint weight matrix, and in combination with the forgetting factor, perform real-time update on the initial judgment threshold to obtain a dynamic discrimination threshold for encoder disconnection faults; S35: According to the dynamic discrimination threshold, calculate the prediction confidence of the deep belief network model.

[0027] In the present invention, step S4 includes the following steps: S41: Use a sliding synchronous window to synchronize the position, speed, current, voltage, and temperature of the steering motor to obtain an operation data stream; S42: Clean, associate, and classify the operation data stream to obtain a loop data stream and an environment data stream; S43: Input the loop data stream into the deep belief network model, and in combination with the dynamic discrimination threshold, give an early warning of the encoder disconnection fault of the steering motor and generate a multi-level decision.

[0028] In the present invention, step S43 includes the following steps, as shown in Table 1: If all loop features in the loop data stream are less than the corresponding dynamic discrimination threshold, it is determined that the encoder is operating normally, and a green warning is displayed; If any loop feature in the loop data stream is greater than or equal to the corresponding dynamic discrimination threshold and lasts for N cycles, a yellow warning of encoder disconnection failure is sent to the administrator; If any two loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence level is greater than or equal to the sub-layer determination threshold, an orange warning of encoder disconnection failure is sent to the administrator; If all loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence level is greater than or equal to the first-layer determination threshold, a red warning of encoder disconnection failure is sent to the administrator.

[0029] Table 1 Refined table of working conditions Operating condition level Trigger condition System response Level 0 No parameter overrun Log recorded, green warning Level 1 Single parameter overrun for 3 cycles Log recorded, yellow warning Level 2 Double parameter overrun + model confidence level > 80% Limited power operation, orange alarm Level 3 Triple parameter overrun + model confidence level > 95% Switch to backup control, red fault In the present invention, step S5 includes the following steps: S51: Perform fitting operations on the temperature and voltage of the environmental data stream to obtain a temperature curve, a voltage curve, and a temperature-voltage fitting formula; S52: Perform disturbance compensation operations on the loop data stream according to the temperature curve in time sequence to obtain a loop temperature-compensated data stream; S53: Perform fluctuation correction operations on the loop temperature-compensated data stream according to the voltage curve in frequency to obtain a loop comprehensive data stream; In this embodiment, a bus voltage feedforward compensation term can be introduced for voltage fluctuation correction; S54: Verify the loop comprehensive data stream according to the temperature-voltage fitting formula, convert it to the frequency domain, obtain the frequency domain characteristics of each loop, and judge the transient working condition of the steering motor.

[0030] In the present invention, step S5 also includes the following steps: If the energy in the low-frequency band of the frequency domain characteristics is concentrated, judge the transient working condition as acceleration or deceleration according to the sign of the current error; If the frequency domain characteristics present characteristic harmonics, judge the transient working condition as straightening or lane changing according to the sign of the current error; If the frequency domain characteristics present a wide-band impact, judge the transient working condition as a load mutation; If the frequency domain characteristics present narrow-band energy concentration, judge the transient working condition as high-frequency increase or decrease according to the sign of the current error; If the frequency domain characteristics present an inherent frequency peak of periodic fluctuation, judge the transient working condition as mechanical resonance.

[0031] Embodiment: Extract the loop commands of the steering motor from the controller: preset position, preset speed, and preset current amplitude; the loop includes a position loop, a speed loop, and a current loop; simulate the loop characteristic data of the steering motor under normal operating conditions: simulated position, simulated speed, and simulated current amplitude; calculate the simulated differences between the loop characteristic data and the corresponding loop commands: position deviation, speed frequency error, and fluctuation amplitude error; based on the simulated differences and the corresponding judgment thresholds, obtain the loop characteristic types of the steering motor under abnormal operating conditions: If the position deviation is greater than the position threshold and the duration reaches the preset value, the loop characteristic type of the abnormal operating condition of the position loop is position integral saturation; if the position deviation is less than or equal to the position threshold but the duration does not reach the preset value, the position loop is in the normal operating condition; If the speed frequency error is greater than the frequency threshold, the loop characteristic type of the abnormal operating condition of the speed loop is speed frequency oscillation; if the speed frequency error is less than or equal to the frequency threshold, the speed loop is in the normal operating condition; If the fluctuation amplitude error is greater than the fluctuation threshold, the loop characteristic type of the abnormal operating condition of the current loop is current overmodulation; if the fluctuation amplitude error is less than or equal to the fluctuation threshold, the current loop is in the normal operating condition; Based on the loop characteristic data of the steering motor under normal operating conditions and the loop characteristic types under abnormal operating conditions, establish a motor simulation data feature library.

[0032] Establish a joint domain measurement layer to perform time-domain analysis and frequency-domain analysis on the loop characteristic data respectively, and obtain the mean value, standard deviation, variance, peak-to-peak value, and spectral energy distribution diagram of each loop; establish a feature decomposition layer to perform wavelet packet decomposition on the loop characteristic data to obtain the data energy entropy of each loop; establish a feature extraction layer to extract features from the loop characteristic data and loop characteristic types to obtain an original feature matrix; establish a multi-dimensional training layer to perform several cross-scale trainings on the original feature matrix in combination with the time domain, frequency domain, and wavelet packet decomposition to obtain a time-frequency domain joint weight matrix: according to the simulated difference and the judgment threshold, the recognition and update layer performs optimization feedback on the time-frequency domain joint weight matrix, and then establishes a deep belief network model of the steering motor.

[0033] Calculate the characteristic deviation degree of each loop according to the covariance matrix combined with the mean value of each loop; obtain the initial judgment threshold according to the standard deviation and mean value of each loop combined with the normal distribution criterion; perform real-time update on the initial judgment threshold according to the sliding window adaptive mechanism and the time-frequency domain joint weight matrix, combined with the forgetting factor, to obtain the dynamic discrimination threshold for encoder disconnection faults; calculate the prediction confidence degree of the deep belief network model according to the dynamic discrimination threshold.

[0034] Synchronize the position, speed, current, voltage, and temperature of the steering motor using a sliding synchronous window to obtain an operating data stream; clean, correlate, and classify the operating data stream to obtain a loop data stream and an environmental data stream; input the loop data stream into a deep belief network model, and combine with a dynamic discrimination threshold to give an early warning of the encoder disconnection fault of the steering motor and generate multi-level decisions: If all loop features in the loop data stream are less than the corresponding dynamic discrimination threshold, it is determined that the encoder is operating normally, and a green early warning is displayed; If any loop feature in the loop data stream is greater than or equal to the corresponding dynamic discrimination threshold and lasts for N cycles, a yellow early warning of the encoder disconnection fault is sent to the administrator; If any two loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence level is greater than or equal to the sub-layer determination threshold, an orange early warning of the encoder disconnection fault is sent to the administrator; If all loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence level is greater than or equal to the first-layer determination threshold, a red early warning of the encoder disconnection fault is sent to the administrator.

[0035] Perform fitting operations on the temperature and voltage of the environmental data stream to obtain a temperature curve, a voltage curve, and a temperature-voltage fitting formula; perform perturbation compensation operations on the loop data stream according to the temperature curve in time sequence to obtain a loop temperature-compensated data stream; perform fluctuation correction operations on the loop temperature-compensated data stream according to the voltage curve in frequency to obtain a loop comprehensive data stream; verify the loop comprehensive data stream according to the temperature-voltage fitting formula and convert it to the frequency domain to obtain the frequency domain characteristics of each loop, and judge the transient working condition of the steering motor: If the energy in the low-frequency band of the frequency domain characteristics is concentrated, judge the transient working condition as acceleration or deceleration according to the current error sign; If the frequency domain characteristics present characteristic harmonics, judge the transient working condition as returning to the correct position or changing lanes according to the current error sign; If the frequency domain characteristics present broadband impulses, judge the transient working condition as a load mutation; If the frequency domain characteristics present narrowband energy concentration, judge the transient working condition as an increase or decrease in high frequency according to the current error sign; If the frequency domain characteristics present inherent frequency spikes with periodic fluctuations, judge the transient working condition as mechanical resonance.

Claims

1. A method for detecting encoder wire breakage of a DC steering motor, characterized in that: It includes the following steps: S1: Simulate the loop characteristic data of the steering motor under full working conditions, and establish a motor simulation data feature library; The full working conditions include normal working conditions and abnormal working conditions; S2: Based on the motor simulation data feature library, establish a deep belief network model for the steering motor; S3: Based on the deep belief network model and combined with the loop characteristic data, generate a dynamic discrimination threshold for encoder disconnection faults; S4: According to the operation data stream of the steering motor and combined with the dynamic discrimination threshold, identify the encoder disconnection fault of the steering motor; The operation data stream includes environmental data stream and loop data stream; S5: Combine the environmental data stream of the steering motor to perform disturbance compensation on the loop data stream, and identify transient working conditions by combining frequency domain characteristics; S6: After the encoder disconnection fault occurs, the controller starts a self-healing execution mechanism to regulate the steering motor, and then updates and optimizes the deep belief network model to complete encoder disconnection detection.

2. The method for detecting encoder wire breakage of a DC steering motor according to claim 1, wherein: The step S1 includes the following steps: S11: Extract the loop instructions of the steering motor from the controller; the loop instructions include preset position, preset speed, and preset current amplitude; The loop includes a position loop, a speed loop, and a current loop; S12: Simulate the loop characteristic data of the steering motor under normal working conditions; the loop characteristic data includes simulated position, simulated speed, and simulated current amplitude; S13: Calculate the simulated difference between the loop characteristic data and the corresponding loop instructions; the simulated difference includes position deviation, speed frequency error, and fluctuation amplitude error; S14: According to the simulated difference and the corresponding judgment threshold, obtain the loop characteristic type of the steering motor under abnormal working conditions.

3. The method for detecting encoder wire breakage of a DC steering motor according to claim 2, wherein: The step S1 also includes the following steps: If the position deviation is greater than the position threshold and the duration reaches the preset value, the loop characteristic type of the abnormal working condition of the position loop is position integral saturation; if the position deviation is less than or equal to the position threshold but the duration does not reach the preset value, the position loop is in normal working condition; If the speed frequency error is greater than the frequency threshold, the loop characteristic type of the abnormal working condition of the speed loop is speed frequency oscillation; if the speed frequency error is less than or equal to the frequency threshold, the speed loop is in normal working condition; If the fluctuation amplitude error is greater than the fluctuation threshold, the loop characteristic type of the abnormal working condition of the current loop is current overmodulation; if the fluctuation amplitude error is less than or equal to the fluctuation threshold, the current loop is in normal working condition; Establish a motor simulation data feature library according to the loop characteristic data of the steering motor under normal working conditions and the loop characteristic type under abnormal working conditions.

4. The method for detecting encoder wire breakage of a DC steering motor according to claim 1, characterized in that: The step S2 includes the following steps: S21: Establish a joint domain calculation layer to perform time-domain analysis and frequency-domain analysis on the loop characteristic data respectively, and obtain the mean value, standard deviation, variance, peak-to-peak value, and spectrum energy distribution diagram of each loop; S22: Establish a feature decomposition layer to perform wavelet packet decomposition on the loop characteristic data, and obtain the data energy entropy of each loop; S23: Establish a feature extraction layer to extract features from the loop characteristic data and the loop characteristic type, and obtain an original feature matrix; S24: Establish a multi-dimensional training layer, and perform several cross-scale trainings on the original feature matrix by combining time domain, frequency domain, and wavelet packet decomposition to obtain a time-frequency domain joint weight matrix: S25: Identify the optimization feedback of the time-frequency domain joint weight matrix by the update layer according to the simulated difference and the judgment threshold, and then establish a deep belief network model for the steering motor.

5. The method for detecting encoder wire breakage of a DC steering motor according to claim 1, characterized in that: The steps in step S3 include the following steps: S31: The deep belief network model trains and corrects the loop feature data to obtain a covariance matrix; S32: Calculate the feature deviation degree of each loop according to the covariance matrix combined with the mean value of each loop; S33: Obtain the initial judgment threshold according to the standard deviation and mean value of each loop combined with the normal distribution criterion; S34: According to the sliding window adaptive mechanism and the time-frequency domain joint weight matrix, and combined with the forgetting factor, the initial judgment threshold is updated in real time to obtain the dynamic discrimination threshold for encoder disconnection faults; S35: Calculate the prediction confidence of the deep belief network model according to the dynamic discrimination threshold.

6. The method for detecting encoder wire breakage of a DC steering motor according to claim 1, wherein: The steps in step S4 include the following steps: S41: Use a sliding synchronous window to synchronize the position, speed, current, voltage and temperature of the steering motor to obtain an operation data stream; S42: Clean, associate and classify the operation data stream to obtain a loop data stream and an environment data stream; S43: Input the loop data stream into the deep belief network model, and combine the dynamic discrimination threshold to give an early warning of the encoder disconnection fault of the steering motor and generate a multi-level decision.

7. The method for detecting encoder wire breakage of a DC steering motor according to claim 5, characterized in that: The steps in step S43 include the following steps: If all loop features in the loop data stream are less than the corresponding dynamic discrimination threshold, it is determined that the encoder is operating normally, and a green early warning is displayed; If any loop feature in the loop data stream is greater than or equal to the corresponding dynamic discrimination threshold and lasts for N cycles, a yellow early warning of encoder disconnection fault is sent to the administrator; If any two loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence is greater than or equal to the sub-layer judgment threshold, an orange early warning of encoder disconnection fault is sent to the administrator; If all loop features in the loop data stream are greater than or equal to the corresponding dynamic discrimination threshold and the prediction confidence is greater than or equal to the first-layer judgment threshold, a red early warning of encoder disconnection fault is sent to the administrator.

8. The method for detecting encoder wire breakage of a DC steering motor according to claim 1, wherein: The steps in step S5 include the following steps: S51: Perform fitting operations on the temperature and voltage of the environment data stream to obtain a temperature curve, a voltage curve and a temperature-voltage fitting formula; S52: Perform disturbance compensation operations on the loop data stream according to the temperature curve in time sequence to obtain a loop temperature compensated data stream; S53: Perform fluctuation correction operations on the loop temperature compensated data stream according to the voltage curve in frequency to obtain a loop comprehensive data stream; S54: Verify the loop comprehensive data stream according to the temperature-voltage fitting formula and convert it to the frequency domain to obtain the frequency domain characteristics of each loop, and judge the transient working condition of the steering motor.

9. The method for detecting encoder wire breakage of a DC steering motor according to claim 1, wherein: The steps in step S5 also include the following steps: If the energy of the low-frequency band of the frequency domain characteristics is concentrated, judge the transient working condition as acceleration or deceleration according to the current error sign; If the frequency domain characteristics present characteristic harmonics, judge the transient working condition as returning to the right or changing lanes according to the current error sign; If the frequency domain characteristics present broadband shocks, judge the transient working condition as a load mutation; If the frequency-domain characteristic shows narrowband energy concentration, then according to the sign of the current error, determine whether the transient condition is a high-frequency increase or decrease; If the frequency-domain characteristic shows a natural frequency peak with periodic fluctuations, then determine that the transient condition is mechanical resonance.

Images