A motor controller bolt loosening online monitoring method and system

Abstract

The application provides a motor controller bolt loosening online monitoring method and system, relates to the motor controller state monitoring and optical fiber sensing technical field, and the method comprises the following steps: acquiring state monitoring data of each measuring point in a monitoring area corresponding to a bolt to be monitored in a motor controller; pre-processing the state monitoring data to obtain strain variation and temperature variation of each measuring point; compensating the strain variation according to the temperature variation to obtain pure mechanical strain of each measuring point; extracting a plurality of characteristic parameters representing bolt pretightening force variation and bolt connection area temperature variation in the monitoring area corresponding to the bolt to be monitored according to the temperature variation and the pure mechanical strain; constructing a bolt loosening online monitoring model based on a machine learning algorithm; and inputting each characteristic parameter into the bolt loosening online monitoring model to output a bolt loosening online monitoring result.

Description

Technical Field

[0001] This invention relates to the field of motor controller status monitoring and fiber optic sensing technology, and in particular to an online monitoring method and system for loose bolts in a motor controller. Background Technology

[0002] With the widespread application of motor controllers in new energy vehicles and other equipment, their internal conductive components, such as copper busbars, are typically connected by bolts. During long-term vehicle operation, the bolted connections are continuously subjected to mechanical vibration, temperature changes, and alternating loads, which can easily lead to a decrease in preload or loosening, thereby affecting the stability of the conductive connection and the operational safety of the motor controller.

[0003] In existing technologies, the detection of bolt connection status in motor controllers typically relies on periodic shutdown inspections, manual retightening, or torque testing. These methods require implementation during vehicle maintenance or equipment downtime, making it difficult to continuously monitor bolt connection status during motor controller operation and potentially failing to detect bolt loosening occurring between adjacent maintenance cycles. Therefore, existing detection methods are insufficient for continuous and timely online monitoring of bolt loosening in motor controllers.

[0004] Furthermore, existing bolt loosening monitoring methods are easily affected by factors such as temperature fluctuations, mechanical vibrations, and changes in operating conditions in actual operating environments. This causes fluctuations in the monitoring signals, making it difficult to accurately distinguish between changes in the state caused by bolt loosening and changes in the operating environment, and also making it difficult to further determine the severity of bolt loosening. Therefore, existing monitoring methods are easily affected by complex operating environments and cannot accurately identify the state and severity of bolt loosening. Summary of the Invention

[0005] To address the technical problems of existing detection methods being unable to continuously and timely monitor the loosening status of motor controller bolts online, and the existing monitoring methods being easily affected by complex operating environments, making it difficult to accurately identify the loosening status and severity of bolts, this invention provides an online monitoring method and system for loosening bolts in motor controllers.

[0006] The technical solution provided by this invention is as follows: In a first aspect, the present invention provides an online monitoring method for loose bolts in a motor controller, comprising: S1: Obtain the status monitoring data of each measuring point in the monitoring area corresponding to the bolt to be monitored in the motor controller; S2: Preprocess the condition monitoring data to obtain the strain change and temperature change at each measuring point; S3: Based on the temperature change, temperature compensation is performed on the strain change to obtain the pure mechanical strain at each measuring point; S4: Based on the temperature change and pure mechanical strain, extract multiple characteristic parameters representing the changes in bolt preload and temperature in the bolt connection area within the monitoring region corresponding to the bolt to be monitored; S5: Construct an online monitoring model for bolt loosening based on machine learning algorithms; S6: Input the various characteristic parameters into the bolt loosening online monitoring model and output the bolt loosening online monitoring results.

[0007] Secondly, the present invention provides an online monitoring system for loose bolts in a motor controller, comprising: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the online monitoring method for loose bolts in a motor controller as described in the first aspect.

[0008] The beneficial effects of the technical solution provided by this invention include: In this embodiment of the invention, by acquiring the status monitoring data of each measuring point within the monitoring area corresponding to the bolt to be monitored in the motor controller, and inputting the extracted feature parameters into the bolt loosening online monitoring model to output the bolt loosening online monitoring results, continuous and timely online monitoring of the bolt loosening state of the motor controller is achieved. By preprocessing the status monitoring data, the strain change and temperature change of each measuring point are obtained. Temperature compensation is performed on the strain change based on the temperature change to obtain pure mechanical strain. Based on the temperature change and pure mechanical strain, multiple feature parameters characterizing the bolt preload change and the temperature change of the bolt connection area are extracted. Then, a bolt loosening online monitoring model based on machine learning algorithm is constructed to handle the influence of complex operating environment and identify the bolt loosening state and its severity. Attached Figure Description

[0009] Figure 1 A flowchart illustrating an online monitoring method for loose bolts in a motor controller, provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of a method for online monitoring of loose bolts in a motor controller, provided by an embodiment of the present invention. Figure 3 A schematic diagram of the arrangement structure of sensing optical fiber in the bolt connection area of ​​the copper busbar of a motor controller, provided by an embodiment of the present invention; Figure 4 This is a schematic diagram of an online monitoring system for loose bolts in a motor controller, provided as an embodiment of the present invention.

[0010] Figure descriptions: 1. Copper busbar; 2. Nut; 3. Ultra-thin annular strain gauge; 4. Bolt; 5. Sensor fiber. Detailed Implementation

[0011] Reference manual attached Figure 1 The diagram shows a flowchart of an online monitoring method for loose bolts in a motor controller provided by an embodiment of the present invention.

[0012] Reference manual attached Figure 2 The diagram shows a structural schematic of an online monitoring method for loose bolts in a motor controller provided by an embodiment of the present invention.

[0013] This invention provides an online monitoring method for loose bolts in a motor controller, the method comprising: S1: Obtain the status monitoring data of each measuring point in the monitoring area corresponding to the bolt to be monitored in the motor controller.

[0014] Optionally, the condition monitoring data includes the core mode temperature sensitivity coefficient, the cladding mode temperature sensitivity coefficient, the core mode strain sensitivity coefficient, the cladding mode strain sensitivity coefficient, the core mode resonant peak wavelength shift, and the cladding mode resonant peak wavelength shift.

[0015] Reference manual attached Figure 3 The diagram shows a schematic of the arrangement of a sensing optical fiber in the bolt connection area of ​​a motor controller copper busbar, according to an embodiment of the present invention.

[0016] like Figure 3 As shown, a micro-groove for embedding and installing the sensing fiber 5 is formed on the upper surface of the copper busbar 1 along its length. The ultra-thin annular strain converter shim 3 is made of an elastic material with a thickness of 1 mm. Its inner diameter matches the diameter of the bolt 4, and its outer diameter is larger than the outer diameter of the nut 2. A spiral micro-groove for embedding and installing the sensing fiber 5 is formed on the cylindrical surface of the ultra-thin annular strain converter shim 3. During installation, the copper busbar 1 is first pre-tightened using the bolt 4, the ultra-thin annular strain converter shim 3, and the nut 2. Then, the sensing fiber 5 is embedded and installed along the micro-groove on the upper surface of the copper busbar 1 and the spiral micro-groove on the cylindrical surface of the ultra-thin annular strain converter shim 3, and filled and fixed with a thermally conductive and insulating adhesive. After the adhesive layer cures, it is polished to make the adhesive surface flat with the surface of the copper busbar 1. After the sensing fiber 5 is installed, the bolt 4 is fully tightened to the required torque. For a motor controller with multiple copper busbars 1 and multiple bolts 4, the same sensing fiber 5 is continuously embedded and installed along each copper busbar 1 and each ultra-thin annular strain gauge 3 in the manner described above, and all bolts 4 are fully tightened after the sensing fiber 5 is installed.

[0017] In one possible implementation, S1 specifically includes sub-steps S101 to S105: S101: Multiple tilted fiber Bragg gratings are set on a single-mode fiber according to the preset measurement point spacing to form a sensing fiber array corresponding to each measurement point.

[0018] S102: Establish the spectral acquisition optical path of the sensing fiber array, acquire the reflection spectrum of each measurement point, and determine the core mode resonance peak and cladding mode resonance peak of each measurement point based on the reflection spectrum.

[0019] S103: Temperature calibration is performed on the core mode resonance peak and the cladding mode resonance peak respectively to obtain the core mode temperature sensitivity coefficient and the cladding mode temperature sensitivity coefficient at each measuring point.

[0020] S104: Under constant temperature conditions, multi-level axial strain is applied to fiber samples from the same batch as the sensing fiber array, and strain calibration is performed on the core mode resonance peak and cladding mode resonance peak respectively to obtain the core mode strain sensitivity coefficient and cladding mode strain sensitivity coefficient at each measuring point.

[0021] S105: Compare the core mode resonant peak wavelength and cladding mode resonant peak wavelength at each measurement point under the initial reference state and the real-time monitoring state to obtain the core mode resonant peak wavelength drift and cladding mode resonant peak wavelength drift at each measurement point.

[0022] Specifically, single-mode fiber is selected, and multiple tilted fiber Bragg gratings are sequentially written onto the single-mode fiber using femtosecond laser direct writing or phase masking according to the preset measurement point spacing. The spacing between adjacent tilted fiber Bragg gratings is set according to the spatial resolution requirements, the tilt angle of each tilted fiber Bragg grating is preferably 4° to 6°, and the wavelength of each tilted fiber Bragg grating is distributed in the C-band, thereby forming a sensing fiber array corresponding to each measurement point.

[0023] Furthermore, a C-band broadband amplified spontaneous emission source is used as the light source. The light output from the source is incident on the sensing fiber array through a circulator, and the reflected light generated by the sensing fiber array is guided out by the circulator to a high-speed fiber optic demodulator or a high-resolution spectrometer to collect the reflection spectra at each measurement point. Based on the reflection spectra at each measurement point, the corresponding core mode resonance peak and the selected cladding mode resonance peak are determined respectively.

[0024] Furthermore, the entire sensing fiber array was placed in a constant temperature chamber, and multiple temperature gradient points were set within the operating temperature range of the sensing fiber array. After reaching thermal stability at each temperature gradient point, the core mode resonant peak wavelength and the selected cladding mode resonant peak wavelength were acquired at each measurement point. Linear fitting was performed on the relationship between the core mode resonant peak wavelength and temperature, as well as the relationship between the cladding mode resonant peak wavelength and temperature, to obtain the core mode temperature sensitivity coefficient and the cladding mode temperature sensitivity coefficient at each measurement point.

[0025] Furthermore, fiber samples from the same batch as the sensing fiber array were cut and fixed on a tensile testing bench. Multiple levels of axial strain were applied to the fiber samples while maintaining a constant temperature. Under each level of axial strain, the wavelength shifts of the core mode resonant peak and the cladding mode resonant peak were obtained. The relationships between the wavelength shifts of the core mode resonant peak and the axial strain, as well as the relationships between the wavelength shifts of the cladding mode resonant peak and the axial strain, were fitted to obtain the core mode strain sensitivity coefficient and the cladding mode strain sensitivity coefficient at each measuring point.

[0026] Finally, the reflection spectra of each measuring point in the sensing fiber array are acquired in real time, and the core mode resonance peak and the target cladding mode resonance peak of each measuring point are tracked using a peak detection algorithm. The wavelength of the core mode resonance peak under real-time monitoring is compared with the corresponding wavelength under the initial reference state, and the wavelength of the cladding mode resonance peak under real-time monitoring is also compared with the corresponding wavelength under the initial reference state, thereby obtaining the wavelength drift of the core mode resonance peak and the wavelength drift of the cladding mode resonance peak at each measuring point.

[0027] In this embodiment of the invention, by embedding sensing optical fibers within the micro-grooves of the motor controller's copper busbars and ultra-thin annular strain conversion pads, and utilizing multiple tilted fiber Bragg gratings to form a sensing optical fiber array corresponding to multiple measurement points, multi-point, continuous monitoring of temperature and strain status in the bolt connection area to be monitored is possible. After embedding and installing the sensing optical fibers, no additional internal space is occupied by the motor controller, and good thermal contact and strain transfer conditions are achieved. Furthermore, multiple copper busbars and multiple bolt connection areas can be connected in series using the same sensing optical fiber, facilitating online centralized monitoring of multiple bolts to be monitored.

[0028] S2: Preprocess the condition monitoring data to obtain the strain change and temperature change at each measuring point.

[0029] In one possible implementation, S2 specifically includes sub-steps S201 and S202: S201: Calculate the strain change at each measuring point based on the core mode temperature sensitivity coefficient, cladding mode temperature sensitivity coefficient, core mode strain sensitivity coefficient, cladding mode strain sensitivity coefficient, core mode resonant peak wavelength drift, and cladding mode resonant peak wavelength drift at each measuring point.

[0030] S202: Calculate the temperature change at each measuring point based on the core mode temperature sensitivity coefficient, cladding mode temperature sensitivity coefficient, core mode strain sensitivity coefficient, cladding mode strain sensitivity coefficient, core mode resonant peak wavelength drift, and cladding mode resonant peak wavelength drift at each measuring point.

[0031] Specifically, the wavelength shifts of both the core mode and cladding mode resonant peaks are caused by both strain and temperature changes, and the temperature and strain sensitivities differ between the core and cladding modes. A two-parameter relationship between temperature and strain is established based on the respective temperature and strain sensitivity coefficients of the core and cladding modes. The two-parameter relationship corresponding to the core mode is multiplied by the temperature sensitivity coefficient of the cladding mode, and vice versa. The results are then subtracted to eliminate the temperature change, thus obtaining the strain change at each measurement point. in, This represents the change in strain at the measuring point. This represents the cladding modulus temperature sensitivity coefficient at the measuring point. This indicates the wavelength shift of the fiber core mode resonant peak at the measurement point. This represents the core-mode temperature sensitivity coefficient at the measuring point. This indicates the wavelength shift of the cladding mode resonant peak at the measurement point. The fiber core modulus strain sensitivity coefficient represents the measurement point. This represents the cladding modulus strain sensitivity coefficient at the measuring point.

[0032] Finally, based on the temperature and strain dual-parameter relationship corresponding to the core mold and the cladding mold, the dual-parameter relationship corresponding to the cladding mold is multiplied by the strain sensitivity coefficient of the core mold, and the results are subtracted to eliminate the strain change, thereby obtaining the temperature change at each measuring point: in, This indicates the temperature change at the measuring point.

[0033] In this embodiment of the invention, by utilizing the differences in temperature sensitivity and strain sensitivity between the core mode and the cladding mode, the wavelength shift of the two types of resonance peaks is jointly processed. This allows the strain change and temperature change at each measurement point to be obtained from the spectral signal that is simultaneously affected by temperature and strain changes. This reduces the interference of the cross-influence of temperature and strain on the monitoring results and provides independent basic monitoring data for subsequent temperature compensation and bolt loosening status identification.

[0034] S3: Based on the temperature change, temperature compensation is performed on the strain change to obtain the pure mechanical strain at each measuring point.

[0035] In one possible implementation, S3 specifically includes sub-steps S301 and S302: S301: Calculate the thermal strain at each measuring point based on the temperature change at each measuring point and the thermal expansion coefficient of the sensing fiber material.

[0036] S302: Subtract the corresponding thermal strain from the strain change at each measuring point to obtain the pure mechanical strain at each measuring point.

[0037] Specifically, the sensing fiber is made of quartz. When the temperature at each measuring point changes, the sensing fiber undergoes corresponding thermal expansion. Based on the linear thermal expansion relationship of the sensing fiber material, the thermal strain at each measuring point is obtained by multiplying the temperature change at each measuring point by the coefficient of thermal expansion of the sensing fiber material. in, Indicates the thermal strain at the measuring point. This represents the coefficient of thermal expansion of the sensing fiber material.

[0038] Finally, since the strain change at each measuring point includes thermal strain caused by temperature change and pure mechanical strain caused by bolt preload change, the pure mechanical strain at each measuring point is obtained by subtracting the corresponding thermal strain from the strain change at each measuring point: in, This represents the purely mechanical strain at the measuring point.

[0039] In this embodiment of the invention, the thermal strain of the sensing fiber due to thermal expansion is calculated based on the temperature change at each measuring point, and the corresponding thermal strain is subtracted from the measured strain change. This separates the strain component caused by temperature change from the mechanical strain component caused by bolt preload change, making the obtained pure mechanical strain more accurately reflect the actual change in the preload of the bolt to be monitored, and avoiding misjudging temperature change as bolt loosening.

[0040] S4: Based on the temperature change and pure mechanical strain, extract multiple characteristic parameters representing the changes in bolt preload and temperature in the bolt connection area within the monitoring region corresponding to the bolt to be monitored.

[0041] Optionally, the characteristic parameters include the average strain, the rate of change of strain, the average temperature, and the rate of change of temperature.

[0042] Specifically, for each bolt to be monitored, within a 1 cm range of the bolt location, the average strain and strain rate of change are extracted based on the pure mechanical strain at each measuring point, and the average temperature and temperature rate of change are extracted based on the temperature change at each measuring point. The average strain, strain rate of change, average temperature, and temperature rate of change are used as multiple characteristic parameters to characterize the changes in bolt preload and temperature changes in the bolt connection area.

[0043] In this embodiment of the invention, within a preset range of the location of the bolt to be monitored, the average strain value, strain change rate, average temperature value, and temperature change rate are extracted respectively. This allows the state of the bolt connection area to be characterized from multiple aspects, including strain level, strain change trend, temperature level, and temperature change trend. This transforms the monitoring information of a single measuring point into characteristic parameters that can comprehensively reflect the changes in bolt preload and temperature changes in the connection area, providing complete input information for subsequent bolt loosening level identification.

[0044] S5: Construct an online monitoring model for bolt loosening based on machine learning algorithms.

[0045] Optionally, the machine learning algorithm is specifically a proportion-dominant ordered logistic regression algorithm.

[0046] In one possible implementation, S5 specifically includes sub-steps S501 to S504: S501: Within the operating temperature range of the motor controller, the average strain, strain change rate, temperature, and temperature change rate of samples corresponding to each loosening level are collected as sample feature parameters to form a training sample set with corresponding loosening levels.

[0047] S502: Calculate the training mean and training standard deviation corresponding to each sample feature parameter, and perform Z-score standardization on each sample feature parameter according to the training mean and training standard deviation to construct a standardized training sample set.

[0048] S503: The standardized training sample set is trained using a proportional dominance ordered logistic regression algorithm to obtain feature regression coefficients and classification thresholds.

[0049] S504: Using the training mean, training standard deviation, feature regression coefficients, and grading thresholds as model parameters, an online monitoring model for bolt loosening is constructed.

[0050] Specifically, four loosening levels—normal, slightly loose, moderately loose, and severely loose—were set on the bolt testing bench, and the sample collection for each loosening level covered the operating temperature range of the motor controller. The average strain, strain change rate, average temperature, and temperature change rate of the samples corresponding to each loosening level were collected. These parameters were then used as sample characteristic parameters, and each characteristic parameter was correlated with its corresponding loosening level to form a training sample set with the corresponding loosening level.

[0051] Furthermore, the training mean and training standard deviation for each sample feature parameter in the training sample set are calculated respectively. Based on the training mean and training standard deviation for each sample feature parameter, the average strain, strain change rate, average temperature, and temperature change rate of the sample in the training sample set are standardized using Z-score. The standardized sample feature parameters are then combined with their corresponding loosening levels to construct a standardized training sample set.

[0052] Furthermore, the standardized training sample set is input into the proportional dominance ordered logistic regression algorithm for training. Based on the relationship between the standardized sample strain average, sample strain change rate, sample temperature average, sample temperature change rate and the corresponding loosening level, the feature regression coefficients corresponding to each sample feature parameter are obtained, as well as three sequentially increasing grading thresholds for classifying the four loosening levels: normal, slightly loose, moderately loose and severely loose.

[0053] Finally, the training mean and standard deviation, feature regression coefficients and grading thresholds corresponding to each sample feature parameter were determined as model parameters. Based on the model parameters, an online monitoring model for bolt loosening based on the proportional advantage ordered logistic regression algorithm was constructed, and the model parameters were embedded into the embedded program.

[0054] In this embodiment of the invention, by collecting training samples corresponding to normal, slightly loose, moderately loose, and severely loose conditions within the operating temperature range of the motor controller, and standardizing the feature parameters of different types of samples, the impact of different dimensions and numerical ranges of the feature parameters on model training can be reduced. A proportional dominance ordered logistic regression algorithm is used for training, which utilizes the sequential nature of the loosening levels to obtain feature regression coefficients and grading thresholds. The model parameters are then embedded in the program for online identification during motor controller operation.

[0055] S6: Input the various characteristic parameters into the bolt loosening online monitoring model and output the bolt loosening online monitoring results.

[0056] Optionally, the online monitoring results for bolt loosening include four loosening levels: normal, slightly loose, moderately loose, and severely loose.

[0057] In one possible implementation, S6 specifically includes sub-steps S601 to S604: S601: Based on the training mean and training standard deviation, Z-score standardization is performed on each feature parameter to obtain multiple standardized feature values.

[0058] S602: Calculate the linear weighted score based on the standardized eigenvalues ​​and eigenre regression coefficients.

[0059] S603: Calculate the three-level cumulative probability using the Sigmoid function based on the linear weighted score and the grading threshold, and calculate the probability of occurrence for each loosening level based on the three-level cumulative probability.

[0060] S604: Compare the occurrence probability of each loosening level, determine the loosening level with the highest occurrence probability as the online monitoring result of bolt loosening and output it.

[0061] Specifically, based on the correspondence between the average strain, strain change rate, average temperature, and temperature change rate and the various sample feature parameters in the training phase, the training mean and training standard deviation determined when constructing the bolt loosening online monitoring model are called respectively. The average strain, strain change rate, average temperature, and temperature change rate corresponding to the current bolt to be monitored are standardized by Z-score to obtain the standardized feature values ​​corresponding to each feature parameter. The standardized feature values ​​are then input into the bolt loosening online monitoring model in the same feature arrangement order as in the model training phase.

[0062] Furthermore, the standardized eigenvalues ​​corresponding to the average strain, strain rate of change, average temperature, and temperature rate of change are multiplied by their respective feature regression coefficients, and the resulting products are summed to form the linear prediction result of the proportional dominance ordered logistic regression algorithm, thus obtaining a linear weighted score used to determine the loosening level of the bolt to be monitored: in, This represents the linearly weighted score. , , and These represent the characteristic regression coefficients corresponding to the average strain, the rate of change of strain, the average temperature, and the rate of change of temperature, respectively. , , and These represent the standardized characteristic values ​​corresponding to the average strain, the rate of change of strain, the average temperature, and the rate of change of temperature, respectively.

[0063] Furthermore, following the order of looseness—normal, slightly loose, moderately loose, and severely loose—the Sigmoid function is used to calculate the cumulative probabilities for three levels based on the relationship between three progressively increasing grading thresholds and linearly weighted scores. The first-level cumulative probability represents the probability that the bolt is in a normal state; the second-level cumulative probability represents the probability that the bolt is in a normal or slightly loose state; and the third-level cumulative probability represents the probability that the bolt is in a normal, slightly loose, or moderately loose state. Subtracting the first-level cumulative probability from the second-level cumulative probability yields the probability of slightly loosening; subtracting the second-level cumulative probability from the third-level cumulative probability yields the probability of moderate loosening. The first-level cumulative probability is used as the probability of normal loosening, and the complement of the third-level cumulative probability is used as the probability of severely loosening. in, , , and These represent the probabilities of the monitored bolts being in normal, slightly loose, moderately loose, and severely loose states, respectively. , and This represents the cumulative probability at level three. , and These represent the grading thresholds between normal and slightly loose, slightly loose and moderately loose, and moderately loose and severely loose, respectively. This represents the Sigmoid function. This represents the input variable of the Sigmoid function. Represents the natural constant.

[0064] Finally, the probability of the monitored bolt being in four loosening levels—normal, slightly loose, moderately loose, and severely loose—is compared. The maximum value among the four probabilities is determined, and the loosening level corresponding to this maximum value is identified as the online monitoring result of the bolt loosening and output.

[0065] In this embodiment of the invention, the real-time feature parameters are standardized according to the training mean and training standard deviation used in the model training phase, which can maintain the consistency of data processing methods between online monitoring data and training samples. By calculating the probability of occurrence of normal, slightly loose, moderately loose, and severely loose bolts through linear weighted scores and three-level cumulative probabilities, and taking the loosening level with the highest probability as the monitoring result, multiple real-time feature parameters can be converted into clear loosening level outputs, while retaining the probability information corresponding to each loosening level, making the process of determining the bolt loosening state clearer.

[0066] In one possible implementation, it also includes: S7: Generate corresponding graded early warning signals based on the online monitoring results of bolt loosening, and report the graded early warning signals to the vehicle so that the vehicle can take handling measures corresponding to the online monitoring results of bolt loosening.

[0067] In one possible implementation, S7 specifically includes sub-steps S701 to S703: S701: When the online monitoring result of bolt loosening is slightly loose, a yellow warning signal is generated and reported to the whole vehicle to remind the bolt to be monitored to be inspected during the next maintenance.

[0068] S702: When the online monitoring result of bolt loosening is moderately loose, an orange warning signal is generated and reported to the whole vehicle so that the whole vehicle can operate with reduced power.

[0069] S703: When the online monitoring result of bolt loosening is severe, a red warning signal is generated and reported to the vehicle to prompt the vehicle to stop immediately.

[0070] In this embodiment of the invention, yellow, orange, and red warning signals are generated according to slight loosening, moderate loosening, and severe loosening, respectively, and reported to the vehicle. This allows the vehicle to take appropriate measures according to the degree of loosening, such as the next maintenance inspection, reduced power operation, and immediate shutdown. This ensures that the warning method matches the degree of bolt loosening, avoids using the same treatment method for different degrees of loosening, and allows for timely measures to be taken before the bolt loosening progresses further.

[0071] Reference manual attached Figure 4 The diagram shows a structural schematic of an online monitoring system for loose bolts in a motor controller provided by an embodiment of the present invention.

[0072] The present invention also provides an online monitoring system 20 for loose bolts in a motor controller, applied to the above-mentioned online monitoring method for loose bolts in a motor controller, comprising: Processor 201.

[0073] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the online monitoring method for loose bolts of the motor controller as described in the method embodiment is implemented.

[0074] The motor controller bolt loosening online monitoring system 20 provided by the present invention can perform the above-mentioned motor controller bolt loosening online monitoring method and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0075] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for online monitoring of loose bolts in a motor controller, characterized in that, include: S1: Obtain the status monitoring data of each measuring point in the monitoring area corresponding to the bolt to be monitored in the motor controller; S2: Preprocess the condition monitoring data to obtain the strain change and temperature change at each of the measuring points; S3: Based on the temperature change, perform temperature compensation on the strain change to obtain the pure mechanical strain at each of the measuring points; S4: Based on the temperature change and the pure mechanical strain, extract multiple characteristic parameters representing the changes in bolt preload and temperature in the bolt connection area within the monitoring region corresponding to the bolt to be monitored; S5: Construct an online monitoring model for bolt loosening based on machine learning algorithms; S6: Input the aforementioned characteristic parameters into the bolt loosening online monitoring model and output the bolt loosening online monitoring results.

2. The online monitoring method for loose bolts in a motor controller according to claim 1, characterized in that: The condition monitoring data includes the core mode temperature sensitivity coefficient, cladding mode temperature sensitivity coefficient, core mode strain sensitivity coefficient, cladding mode strain sensitivity coefficient, core mode resonant peak wavelength shift, and cladding mode resonant peak wavelength shift; S1 specifically includes: S101: Multiple tilted fiber Bragg gratings are set on a single-mode fiber according to a preset measurement point spacing to form a sensing fiber array corresponding to each measurement point. S102: Establish the spectral acquisition optical path of the sensing fiber array, acquire the reflection spectrum of each measurement point, and determine the core mode resonance peak and cladding mode resonance peak of each measurement point based on the reflection spectrum. S103: Temperature calibration is performed on the core mode resonance peak and the cladding mode resonance peak respectively to obtain the core mode temperature sensitivity coefficient and the cladding mode temperature sensitivity coefficient at each of the measurement points; S104: Under constant temperature conditions, apply multi-level axial strain to fiber samples from the same batch as the sensing fiber array, and perform strain calibration on the core mode resonance peak and the cladding mode resonance peak respectively to obtain the core mode strain sensitivity coefficient and the cladding mode strain sensitivity coefficient at each measurement point. S105: Compare the core mode resonant peak wavelength and cladding mode resonant peak wavelength of each of the measurement points under the initial reference state and the real-time monitoring state, respectively, to obtain the core mode resonant peak wavelength drift and cladding mode resonant peak wavelength drift of each of the measurement points.

3. The online monitoring method for loose bolts in a motor controller according to claim 2, characterized in that: S2 specifically includes: S201: Calculate the strain change at each of the measuring points based on the core mode temperature sensitivity coefficient, cladding mode temperature sensitivity coefficient, core mode strain sensitivity coefficient, cladding mode strain sensitivity coefficient, core mode resonant peak wavelength drift, and cladding mode resonant peak wavelength drift at each of the measuring points. S202: Calculate the temperature change at each of the measuring points based on the core mode temperature sensitivity coefficient, cladding mode temperature sensitivity coefficient, core mode strain sensitivity coefficient, cladding mode strain sensitivity coefficient, core mode resonant peak wavelength drift, and cladding mode resonant peak wavelength drift at each of the measuring points.

4. The online monitoring method for loose bolts in a motor controller according to claim 1, characterized in that: S3 specifically includes: S301: Calculate the thermal strain at each of the measuring points based on the temperature change at each measuring point and the thermal expansion coefficient of the sensing fiber material; S302: Subtract the corresponding thermal strain from the strain change at each of the measuring points to obtain the pure mechanical strain at each of the measuring points.

5. The online monitoring method for loose bolts in a motor controller according to claim 1, characterized in that: The characteristic parameters include average strain, strain change rate, average temperature, and temperature change rate; the machine learning algorithm is specifically a proportional dominance ordered logistic regression algorithm; the online monitoring results of bolt loosening include four loosening levels: normal, slightly loose, moderately loose, and severely loose.

6. The online monitoring method for loose bolts in a motor controller according to claim 5, characterized in that: S5 specifically includes: S501: Within the operating temperature range of the motor controller, the average strain value, strain change rate, average temperature value, and temperature change rate of the samples corresponding to each loosening level are collected as sample feature parameters to form a training sample set with corresponding loosening levels. S502: Calculate the training mean and training standard deviation corresponding to each of the sample feature parameters, and perform Z-score standardization on each of the sample feature parameters according to the training mean and the training standard deviation to construct a standardized training sample set; S503: The standardized training sample set is trained using the proportional dominance ordered logistic regression algorithm to obtain feature regression coefficients and classification thresholds; S504: Using the training mean, the training standard deviation, the feature regression coefficient, and the grading threshold as model parameters, construct the bolt loosening online monitoring model.

7. The online monitoring method for loose bolts in a motor controller according to claim 6, characterized in that: S6 specifically includes: S601: Based on the training mean and the training standard deviation, each of the feature parameters is Z-score standardized to obtain multiple standardized feature values; S602: Calculate the linear weighted score based on the standardized eigenvalues ​​and regression coefficients of each item; S603: Based on the linear weighted score and the grading threshold, calculate the three-level cumulative probability using the Sigmoid function, and calculate the probability of occurrence of each loosening level based on the three-level cumulative probability; S604: Compare the occurrence probability of each loosening level, determine the loosening level with the highest occurrence probability as the online monitoring result of bolt loosening and output it.

8. The online monitoring method for loose bolts in a motor controller according to claim 5, characterized in that: Also includes: S7: Generate a corresponding graded early warning signal based on the online monitoring results of bolt loosening, and report the graded early warning signal to the vehicle so that the vehicle can take handling measures corresponding to the online monitoring results of bolt loosening.

9. The online monitoring method for loose bolts in a motor controller according to claim 8, characterized in that: Specifically, S7 includes: S701: When the online monitoring result of the bolt loosening is slightly loose, a yellow warning signal is generated and the yellow warning signal is reported to the vehicle to remind the monitoring bolt to be inspected during the next maintenance. S702: When the online monitoring result of the bolt loosening is moderate loosening, an orange warning signal is generated and the orange warning signal is reported to the vehicle so that the vehicle can operate with reduced power. S703: When the online monitoring result of the bolt loosening is that it is severely loose, a red warning signal is generated and the red warning signal is reported to the vehicle to prompt the vehicle to stop immediately.

10. An online monitoring system for loose bolts in a motor controller, characterized in that, include: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the online monitoring method for loose bolts in a motor controller as described in any one of claims 1 to 9.

Images