Automobile generator performance detection method based on intelligent sensing system

Abstract

The invention relates to the technical field of new energy automobile generator performance detection, and discloses an automobile generator performance detection method based on an intelligent sensing system, and the method comprises the steps: collecting and constructing a long-term dual-scale magnetic field feature matrix; extracting a performance trend characteristic matrix; outputting a continuous performance state scoring index; performing trend prediction analysis on the performance state scoring index; and outputting a generator stator performance detection report. In the prior art, manual regular inspection and overall electrical parameter detection are relied on, and especially in the early stage of slow attenuation and weak abnormal state change of the internal performance of a generator stator winding, timely performance detection cannot be realized. Due to the fact that high-precision detection of the long-term performance state of the automobile generator stator winding is achieved through the multi-scale magnetic field data collection and fusion analysis method, the problem that the winding is damaged due to the fact that the winding is found only when the performance of the winding is seriously degraded is solved, and the operation safety of an automobile generator system is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of performance detection of new energy vehicle generators, and particularly relates to a method for detecting the performance of vehicle generators based on an intelligent sensing system. Background Art

[0002] With the development of the automotive industry, as a key device for vehicle power supply, the operating state of the new energy vehicle generator is directly related to the safety and reliability of the whole vehicle. At present, the performance detection and maintenance of the stator winding of the new energy vehicle generator mainly rely on manual regular inspections or simple detection methods based on overall electrical parameters (such as output voltage and current). These methods generally have the deficiencies of low monitoring efficiency, limited detection accuracy, and difficulty in perceiving early state changes. Specifically, the period of manual regular inspections is usually long, and it is difficult to detect the early signs of the slow decay of the internal insulation performance of the new energy vehicle generator in a timely manner. Once this slow decay enters the obvious stage, it often means that the insulating material has undergone irreversible aging or local damage, which will bring serious equipment safety hazards. In addition, due to the complex structure of the stator winding of the new energy vehicle generator and the harsh operating environment (such as long-term high temperature, vibration, and frequent load changes), the internal winding insulating material is extremely prone to local aging and local performance degradation. The magnetic field characteristic changes shown in the initial stage of this local degradation are weak and hidden, and it is very difficult for traditional detection means to sensitively capture them. Therefore, there is an urgent need to propose a new performance detection method that can still accurately and timely achieve long-term state monitoring, trend prediction, and performance evaluation in the initial stage of the slow decay and weak abnormal state changes of the stator winding performance of the generator, and further realize dynamic early maintenance decision recommendation, so as to avoid serious damage and accident risks caused by the failure to detect the deterioration of the winding performance in time, and effectively ensure the long-term stable operation of the new energy vehicle generator system. Summary of the Invention

[0003] Aiming at the above-mentioned existing technical deficiencies, the purpose of the present invention is to propose a method for detecting the performance of vehicle generators based on an intelligent sensing system, aiming to solve the technical problem in the prior art that relies on manual regular inspections and overall electrical parameter detections, especially in the early stage of the slow decay of the internal performance of the stator winding of the generator and weak abnormal state changes, and cannot achieve timely performance detection.

[0004] To solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a method for detecting the performance of vehicle generators based on an intelligent sensing system,

[0005] The method for detecting the performance of vehicle generators based on an intelligent sensing system includes:

[0006] Step S10: With the center of the stator winding core of the automotive generator as the center and a preset distance Δr, construct a dual-scale Hall magnetic field sensor array at positions with radii of R and R + Δr on the outer periphery of the stator winding core respectively. Continuously and synchronously collect the magnetic field intensity data of each sensor in real time to form a long-term dual-scale magnetic field feature matrix;

[0007] Step S20: Use the improved empirical mode decomposition and singular spectrum analysis fusion method to extract features from the long-term dual-scale magnetic field feature matrix to obtain a performance trend feature matrix;

[0008] Step S30: With the performance trend feature matrix obtained in Step S20 as the input, construct a spatial frequency domain performance evaluation model and output a continuous performance status scoring index;

[0009] Step S40: Analyze the change of the performance status of the continuous performance status scoring index through the circular normal distribution and time series trend fusion method to obtain the performance status change trend;

[0010] Step S50: According to the performance status change trend obtained in Step S40, combine reinforcement learning to construct a dynamic maintenance decision recommendation model and output a generator stator performance detection report.

[0011] Preferably, in Step S10, 6 Hall magnetic field sensors are arranged at intervals of 60° in the inner circle with a radius of R, and 3 Hall magnetic field sensors are arranged at intervals of 120° in the outer circle with a radius of R + Δr, where Δr is 10 mm to 15 mm; the magnetic sensitive axis directions of all Hall magnetic field sensors point radially to the center of the circle; the long-term dual-scale magnetic field feature matrix is expressed as:

[0012]

[0013] where B R (·, t) is the magnetic field feature at the position with a radius of R at time t, and B R+Δr (·, t) is the magnetic field feature at the position with a radius of R + Δr at time t.

[0014] Preferably, in Step S20, the steps of using the improved empirical mode decomposition and singular spectrum analysis fusion method to extract features from the long-term dual-scale magnetic field feature matrix to obtain a performance trend feature matrix specifically include:

[0015] Step S201: Use the improved empirical mode decomposition method with the introduction of spatial period constraint conditions to perform mode decomposition processing on the long-term dual-scale magnetic field feature matrix to obtain effective intrinsic mode function components reflecting the long-term performance degradation trend;

[0016] Step S202: Further perform singular value decomposition processing on the effective intrinsic mode function components using the singular spectrum analysis fusion method to obtain a performance trend feature matrix reflecting the long-term degradation trend of the stator winding performance

[0017] Preferably, in step S30, taking the performance trend feature matrix obtained in step S20 as the input, the steps of constructing the spatial frequency domain performance evaluation model specifically include:

[0018] Construct a spatial domain attention mechanism layer: used to analyze the correlation of the performance trend feature matrix at different sensor spatial positions to obtain spatial weights;

[0019] Construct a frequency domain attention mechanism layer: used to perform Fourier spectrum transformation on the performance trend feature matrix, automatically capture the frequency domain characteristics of the magnetic field performance changes, and obtain frequency domain weights;

[0020] Construct a Transformer fusion layer: used to fuse the above spatial weights and frequency domain weights and form the final performance state evaluation features through an adaptive weighting method;

[0021] Construct an output layer: used to output a continuous performance state score index PHI according to the performance state evaluation features. The range of PHI is from 0 to 1, which is used to represent the current performance health state of the stator winding. A value close to 1 indicates a good performance state, and a value close to 0 indicates approaching performance degradation and failure.

[0022] Preferably, in step S30, the spatial frequency domain performance evaluation model is trained using historical measured data and simulated magnetic field performance data, including: pre-training the spatial frequency domain performance evaluation model using simulation data with a transfer learning strategy, and fine-tuning the model parameters using measured data.

[0023] Preferably, in step S40, the steps of analyzing the performance state change of the continuous performance state score index through the circular normal distribution and time series trend fusion method to obtain the performance state change trend specifically include:

[0024] Step S401: Based on the continuous performance state score index output in step S30, construct a long-term performance state sequence, and use the circular normal distribution Von-Mises to perform probability density interpolation fitting on the long-term performance state sequence. The specific formula is:

[0025]

[0026] Among them, θ is the angular value representing the mapping of the performance status scoring index to the circular space; μ is the mean parameter of the circular distribution; k is the concentration parameter of the circular distribution; I0(k) is the zero-order modified Bessel function; f(θ|μ,k) is the probability density interpolation fitting function;

[0027] Step S402: Use maximum likelihood estimation to solve the parameters μ and k of the circular normal distribution Von-Mises in Step S401 to obtain an optimized probability density fitting model;

[0028] Step S403: Use the optimized probability density fitting model obtained in Step S402 to predict the trend of the performance status scoring index within a preset future time period through time series prediction to obtain a future performance change trend curve;

[0029] Step S404: Calculate the performance attenuation degree and the corresponding remaining life confidence interval of the automotive generator stator winding within a future specified period according to the future performance change trend curve to obtain the performance status change trend.

[0030] Preferably, in Step S50, the steps of constructing a dynamic maintenance decision recommendation model based on the performance status change trend obtained in Step S40 and outputting a generator stator performance detection report specifically include:

[0031] Step S501: Obtain historical maintenance operation record data and historical maintenance operation effect data, use the performance status change trend as the input feature of the reinforcement learning state, and define the system state vector S in combination with the historical maintenance operation record data;

[0032] Step S502: Define the maintenance action space A based on the historical maintenance operation record data, specifically including:

[0033] A1: Do not perform maintenance for the time being and continue monitoring; A2: Local insulation cleaning and maintenance; A3: Replacement of local insulation components; A4: Local repair of the winding; A5: Overall replacement of the winding;

[0034] Step S503: Construct a reward function R(S,A) according to the maintenance action space A in combination with the historical maintenance operation effect data:

[0035] R(S,A) = w1ΔPHI recovery (A) - w2Cost(A) - w3Risk(S)

[0036] Among them, ΔPHI recovery(A) is the expected performance score recovery value after performing action A; Cost(A) is the cost required to maintain action A; Risk(S) is the risk of the future performance state when maintenance is not performed; w1, w2, and w3 are adjustable weight parameters used to adjust the importance ratios of performance improvement, cost, and risk respectively;

[0037] Step S504: Construct a dynamic maintenance decision recommendation model according to the reward function R(S,A). The dynamic maintenance decision recommendation model specifically includes:

[0038] Input layer: Receives the system state vector;

[0039] Hidden layer: Adopts a two-stream network structure, including:

[0040] Stream A: Contains 2 fully connected layers, with the number of neurons in each layer being 64 and 32 respectively. The activation function takes ReLU and is used to analyze the performance state trend;

[0041] Stream B: Contains 2 fully connected layers, with the number of neurons in each layer being 64 and 32 respectively. The activation function takes ReLU and is used to estimate the expected effect of each maintenance action;

[0042] Fusion layer: Concatenates the feature vectors output by Stream A and Stream B to form a joint state-action evaluation vector;

[0043] Output layer: The output layer is a fully connected layer used to calculate the action value Q value of each maintenance action, generate a performance detection report of the generator stator and output it.

[0044] The present invention also provides an automotive generator performance detection system based on an intelligent sensing system, including:

[0045] Magnetic field data acquisition module, which is used to construct a dual-scale Hall magnetic field sensor array at positions with radii of R and R + Δr respectively on the outer periphery of the stator winding core with the center of the stator winding core of the automotive generator as the center of the circle, continuously and real-time synchronously collect the magnetic field intensity data of each sensor, and form a long-term dual-scale magnetic field feature matrix;

[0046] Performance feature extraction module, which is used to extract features from the long-term dual-scale magnetic field feature matrix by using an improved empirical mode decomposition and singular spectrum analysis fusion method to obtain a performance trend feature matrix;

[0047] State score calculation module, which is used to construct a spatial frequency domain performance evaluation model with the performance trend feature matrix obtained in step S20 as the input and output a continuous performance state score index;

[0048] A state change analysis module, which is used to analyze the performance state change of continuous performance state scoring indexes through the circular normal distribution and time series trend fusion method, and obtain the performance state change trend;

[0049] A detection report generation module, which is used to construct a dynamic maintenance decision recommendation model based on reinforcement learning in combination with the performance state change trend obtained in step S40, and output a generator stator performance detection report.

[0050] The present invention also provides a computer program product, including an automotive generator performance detection program based on an intelligent sensing system. When the automotive generator performance detection program based on the intelligent sensing system is executed by a processor, the automotive generator performance detection method based on the intelligent sensing system described above is implemented.

[0051] The beneficial effects of the present invention are as follows: The present invention realizes the high-precision detection of the long-term continuous performance state of the stator winding of a new energy vehicle generator through an intelligent sensing system, can timely and accurately capture weak magnetic field abnormal changes in the early stage of slow attenuation of the winding performance, avoids the risk of damage caused by discovering after the winding performance deteriorates severely, and significantly improves the operation safety of the new energy vehicle generator.

[0052] In addition, by combining the intelligent sensing system with the trend prediction analysis method, it is possible to accurately predict the future performance change trend of the stator winding of a new energy vehicle generator, and construct a dynamic maintenance decision recommendation model based on reinforcement learning, realizing targeted early maintenance guidance, and significantly improving the accuracy and reliability of generator system maintenance. Description of the Drawings

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0054] Figure 1 It is a schematic flowchart of the first embodiment of an automotive generator performance detection method based on an intelligent sensing system of the present invention.

[0055] Figure 2 It is a schematic diagram of the device of an automotive generator performance detection method based on an intelligent sensing system of the present invention. Detailed Embodiments

[0056] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0057] Embodiment 1: As Figure 1 shown, it is a schematic flowchart of the first embodiment of the method for detecting the performance of an automotive generator based on an intelligent sensing system of the present invention, and the first embodiment of the method for detecting the performance of an automotive generator based on an intelligent sensing system of the present invention is proposed.

[0058] In the first embodiment, the method for detecting the performance of an automotive generator based on an intelligent sensing system includes:

[0059] Step S10: Taking the center of the iron core of the stator winding of the automotive generator as the center of the circle, a preset distance Δr is set, and a dual-scale Hall magnetic field sensor array is constructed at positions with radii of R and R + Δr on the outer periphery of the iron core of the stator winding respectively. Continuously and real-time synchronously collect the magnetic field intensity data of each sensor to form a long-term dual-scale magnetic field feature matrix;

[0060] It should be noted that in step S10, 6 Hall magnetic field sensors are arranged at intervals of 60° in the inner circle with a radius of R, and 3 Hall magnetic field sensors are arranged at intervals of 120° in the outer circle with a radius of R + Δr, and Δr is 10 mm to 15 mm; the magnetic sensitive axis directions of all Hall magnetic field sensors point radially towards the center of the circle; the long-term dual-scale magnetic field feature matrix is expressed as:

[0061]

[0062] where B R (·, t) is the magnetic field feature at the position with a radius of R at time t, and B R+Δr (·, t) is the magnetic field feature at the position with a radius of R + Δr at time t.

[0063] It should be understood that through the dual-scale Hall magnetic field sensor array constructed by the intelligent sensing system, the spatial resolution of magnetic field data acquisition is effectively increased, and the detection sensitivity of weak magnetic field anomalies inside the stator winding of the generator can be significantly improved, thus realizing a more refined and accurate long-term continuous performance monitoring effect. The described dual-scale magnetic field feature matrix increases the data dimension of magnetic field detection through the fusion of inner and outer circle magnetic field data, exceeds the traditional single-scale sensor layout method in terms of both spatial scale and sensitivity, and can more comprehensively capture the weak change trend characteristics of the magnetic field inside the winding, significantly improving the reliability and stability of long-term performance state detection.

[0064] For example, in actual experiments, when the local insulation performance of the stator winding of an automotive generator is slowly degrading in the initial stage, the amplitude of the magnetic field change detected by a traditional single-scale magnetic field sensor array is only between 0.01 and 0.02 mT, making it difficult to achieve accurate early detection. However, when using the dual-scale Hall magnetic field sensor array of this embodiment, the amplitude of the magnetic field change detected under the same conditions can reach 0.05 to 0.08 mT, enabling the identification of the degradation trend of the winding performance about 3 months in advance, significantly improving the sensitivity and reliability of early detection.

[0065] Step S20: Extract features from the long-term dual-scale magnetic field feature matrix using an improved empirical mode decomposition and singular spectrum analysis fusion method to obtain a performance trend feature matrix.

[0066] It should be noted that in step S20, the step of extracting features from the long-term dual-scale magnetic field feature matrix using an improved empirical mode decomposition and singular spectrum analysis fusion method to obtain a performance trend feature matrix specifically includes:

[0067] Step S201: Perform modal decomposition on the long-term dual-scale magnetic field feature matrix using an improved empirical mode decomposition method with spatial period constraint conditions to obtain effective intrinsic mode function components reflecting the long-term performance degradation trend.

[0068] Step S202: Further perform singular value decomposition on the effective intrinsic mode function components using the singular spectrum analysis fusion method to obtain a performance trend feature matrix reflecting the long-term degradation trend of the stator winding performance.

[0069] It should be understood that the improved method with spatial period constraint solves the mode mixing problem of the traditional EMD algorithm, and at the same time, the SSA algorithm can more accurately capture and extract the subtle change features in the long-term trend signal. Therefore, it effectively improves the accuracy and sensitivity of the performance trend feature extraction of the automotive generator stator winding. Compared with the traditional single method based only on EMD or only on SSA, it can more accurately capture and reconstruct the trend signal features of long-term performance changes, thus significantly improving the signal-to-noise ratio and stability of the feature signals in long-term monitoring data, providing a more accurate and reliable data basis for the accurate assessment of subsequent performance states and trend prediction.

[0070] For example, in actual experiments, for the long-term trend signal features extracted separately using the traditional EMD method, the average signal-to-noise ratio (SNR) of the performance degradation trend is only about 12 dB. However, for the trend feature matrix extracted using the improved EMD and SSA fusion method of this embodiment, the signal-to-noise ratio can be increased to more than 28 dB, greatly improving the detection sensitivity and reliability of the performance trend features, and increasing the accuracy of subsequent performance trend prediction by more than 20%.

[0071] Step S30: Using the performance trend feature matrix obtained in step S20 as input, construct a spatial frequency domain performance evaluation model, and output a continuous performance status scoring index;

[0072] It should be noted that in step S30, the steps of constructing a spatial frequency domain performance evaluation model with the performance trend feature matrix obtained in step S20 as input specifically include:

[0073] Construct a spatial domain attention mechanism layer: used to analyze the correlation of the performance trend feature matrix at different sensor spatial positions and obtain spatial weights;

[0074] Construct a frequency domain attention mechanism layer: used to perform Fourier spectrum transformation on the performance trend feature matrix, automatically capture the frequency domain characteristics of the magnetic field performance change, and obtain frequency domain weights;

[0075] Construct a Transformer fusion layer: used to fuse the above spatial weights and frequency domain weights and form the final performance status evaluation feature through an adaptive weighting method;

[0076] Construct an output layer: used to output a continuous performance status scoring index PHI according to the performance status evaluation feature. The range of PHI is from 0 to 1, which is used to represent the current performance health status of the stator winding. A value close to 1 indicates a good performance status, and a value close to 0 indicates approaching performance degradation and failure.

[0077] It should be understood that the spatial frequency domain performance evaluation model creatively adopts a dual attention mechanism to simultaneously mine the key information of the magnetic field trend signal from two dimensions of spatial position and spectral characteristics, and then deeply fuses the features of these two dimensions through the Transformer fusion layer, effectively improving the accuracy and robustness of the performance status evaluation of the generator stator winding. Compared with traditional single feature extraction methods or conventional machine learning classification methods, it can make full use of the complementary features of the magnetic field performance data in the spatial dimension and the spectral dimension, effectively avoiding the defects of insufficient information extraction in the single dimension and being susceptible to noise interference, and significantly improving the recognition accuracy and stability of the performance status.

[0078] For example, in actual experiments, when using a traditional convolutional neural network (CNN) to classify the performance status of magnetic field trend data, the average accuracy of the model is only about 83%. On the same data set, the correlation between the performance status scoring index PHI of the spatial frequency domain performance evaluation model proposed in this embodiment and the actual performance degradation situation reaches more than 95%, and the overall accuracy of the performance status evaluation is improved by more than 12%, effectively realizing the accurate and stable evaluation of the performance status.

[0079] Step S40: Analyze the change in the performance state of the continuous performance state scoring index through the method of fusing circular normal distribution and time series trend to obtain the trend of performance state change;

[0080] It should be noted that in step S40, the steps of analyzing the change in the performance state of the continuous performance state scoring index through the method of fusing circular normal distribution and time series trend to obtain the trend of performance state change specifically include:

[0081] Step S401: Construct a long-term performance state sequence based on the continuous performance state scoring index output in step S30, and use the circular normal distribution Von-Mises to perform probability density interpolation fitting on the long-term performance state sequence. The specific formula is:

[0082]

[0083] where θ is the angle value representing the performance state scoring index mapped to the circular space; μ is the mean parameter of the circular distribution; k is the concentration parameter of the circular distribution; I0(k) is the zero-order modified Bessel function; f(θ|μ,k) is the probability density interpolation fitting function;

[0084] Step S402: Use the maximum likelihood estimation to solve the parameters μ and k of the circular normal distribution Von-Mises in step S401 to obtain an optimized probability density fitting model;

[0085] Step S403: Use the optimized probability density fitting model obtained in step S402 to predict the trend of the performance state scoring index within a preset future time period through the time series prediction method to obtain the future performance change trend curve;

[0086] Step S404: Calculate the performance attenuation degree and the corresponding remaining life confidence interval of the automotive generator stator winding within a specified future period according to the future performance change trend curve to obtain the trend of performance state change.

[0087] It should be understood that in this embodiment, a method of fusing a circular normal distribution Von-Mises distribution model with a time series prediction algorithm is creatively adopted, which accurately realizes the joint modeling analysis of the periodicity and trend of the long-term sequence of the performance state, breaks through the limitation that the traditional single linear trend prediction method cannot capture the periodic change characteristics, and makes the trend analysis of the performance state more conform to the actual operation characteristics of the automotive generator stator winding. In this embodiment, the circular normal distribution is used to perform probability density interpolation modeling on the performance state sequence, which can effectively capture the periodic characteristics of the long-term change of the performance state, and accurately analyze the trend of performance decline through the time series prediction model, so as to obtain clear and detailed future performance trend prediction results. Compared with the traditional method, the accuracy, stability and reliability of the trend prediction are significantly improved, providing a more reliable support basis for the generator maintenance decision-making.

[0088] For example, in actual experimental tests, when using the traditional single ARIMA method to predict the future trend of the performance state index sequence, the average prediction error reaches 10% - 15%. However, in this embodiment, a method of fusing the circular normal distribution with ARIMA is creatively adopted, and the prediction error can be significantly reduced to 4% - 6% on the same data set. The average correlation coefficient between the predicted value of the performance state score index PHI and the actual monitored value within the next 30 days exceeds 0.95, and the prediction accuracy of the performance decline trend is improved by more than double, greatly improving the reliability of the generator performance state trend prediction and the accuracy of the early maintenance guidance.

[0089] Step S50: According to the performance state change trend obtained in step S40, combine reinforcement learning to construct a dynamic maintenance decision recommendation model, and output a generator stator performance detection report.

[0090] It should be noted that in step S50, the steps of constructing a dynamic maintenance decision recommendation model and outputting a generator stator performance detection report according to the performance state change trend obtained in step S40 and combining reinforcement learning specifically include:

[0091] Step S501: Obtain historical maintenance operation record data and historical maintenance operation effect data, use the performance state change trend as the input feature of the reinforcement learning state, and define the system state vector S in combination with the historical maintenance operation record data;

[0092] Step S502: Define the maintenance action space A based on the historical maintenance operation record data, specifically including:

[0093] A1: Do not perform maintenance for the time being and continue to monitor; A2: Local insulation cleaning and maintenance; A3: Replacement of local insulation components; A4: Local repair of the winding; A5: Overall replacement of the winding;

[0094] Step S503: Construct a reward function R(S,A) based on the maintenance action space A and in combination with the historical maintenance operation effect data:

[0095] R(S,A) = w1ΔPHI recovery (A) - w2Cost(A) - w3Risk(S)

[0096] where ΔPHI recovery (A) is the expected performance score recovery value after executing action A; Cost(A) is the cost required for maintenance action A; Risk(S) is the risk of the future performance state without performing maintenance; w1, w2, and w3 are adjustable weight parameters, which are respectively used to adjust the importance ratios of performance improvement, cost, and risk;

[0097] Step S504: Construct a dynamic maintenance decision recommendation model according to the reward function R(S,A). The dynamic maintenance decision recommendation model specifically includes:

[0098] Input layer: Receive the system state vector;

[0099] Hidden layer: Adopt a two-stream network structure, including:

[0100] Stream A: It contains 2 fully connected layers, with the number of neurons in each layer being 64 and 32 respectively, and the activation function takes ReLU, which is used to analyze the performance state trend;

[0101] Stream B: It contains 2 fully connected layers, with the number of neurons in each layer being 64 and 32 respectively, and the activation function takes ReLU, which is used to estimate the expected effect of each maintenance action;

[0102] Fusion layer: Feature splice the feature vectors output by Stream A and Stream B to form a joint state-action evaluation vector;

[0103] Output layer: The output layer is a fully connected layer, which is used to calculate the action value Q value of each maintenance action, generate a performance detection report of the generator stator and output it.

[0104] It should be understood that in this embodiment, a dynamic maintenance decision recommendation model is creatively constructed in combination with reinforcement learning. This model makes full use of historical maintenance operation data and real-time performance trend state data, can dynamically and accurately evaluate and recommend the best maintenance actions for the automotive generator stator winding, realizes the early response to the risk of future performance deterioration, and greatly reduces the risk of unplanned failures of the generator. Compared with the traditional single-stream structure, the two-stream network structure reinforcement learning model constructed in this embodiment can more finely model and deeply integrate the generator performance state trend characteristics and maintenance action characteristics respectively, so as to more accurately evaluate the actual effect of maintenance actions on performance recovery, and effectively improve the reliability and economy of maintenance decisions.

[0105] For example, in actual test cases, the matching degree between the maintenance actions recommended by the traditional static maintenance decision-making method and the actual performance recovery effect is only about 72% on average; after adopting the reinforcement learning dynamic maintenance decision-making recommendation model of this embodiment, the matching degree between the maintenance actions and the actual performance recovery effect is increased to over 90%, and the maintenance cost is reduced by about 25% on average, which fully proves the advantages of the method of the present invention in dynamic maintenance decision-making recommendation.

[0106] Embodiment 2: In addition, a performance detection system for an automotive generator based on an intelligent sensing system provided by the present invention adopts the performance detection method for an automotive generator based on an intelligent sensing system in the above embodiment, and can solve the technical problem of performance detection for an automotive generator based on an intelligent sensing system. Compared with the prior art, the beneficial effects of the performance detection system for an automotive generator based on an intelligent sensing system provided by the present invention are the same as those of the performance detection method for an automotive generator based on an intelligent sensing system provided in the above embodiment, and other technical features in the performance detection system for an automotive generator based on an intelligent sensing system are the same as the features disclosed in the method of the above embodiment, and will not be elaborated here.

[0107] Embodiment 3: The present invention provides a performance detection device for an automotive generator based on an intelligent sensing system. Please refer to Figure 2, An automotive generator performance detection device based on an intelligent sensing system includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method for detecting the performance of an automotive generator based on an intelligent sensing system in the first embodiment above. An automotive generator performance detection device in an embodiment of the present invention may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistant), PADs (Portable Application Description), PMPs (Portable Media Player), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. An automotive generator performance detection device based on an intelligent sensing system is merely an example and should not impose any limitations on the functions and usage scope of embodiments of the present invention. An automotive generator performance detection device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM: Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of an automotive generator performance detection device are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow an automotive generator performance detection device to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows an automotive generator performance detection device having various systems, it should be understood that it is not required to implement or have all the shown systems. Instead, more or fewer systems may be implemented or had.

[0108] Embodiment 4: The present invention further provides a computer program product, including a computer program, which when executed by a processor, implements the steps of a method for detecting the performance of an automotive generator based on an intelligent sensing system as described above. The computer program product provided by the present invention can solve the technical problem of detecting the performance of an automotive generator based on an intelligent sensing system. Compared with the prior art, the beneficial effects of the computer program product provided by the present invention are the same as those of the method for detecting the performance of an automotive generator based on an intelligent sensing system provided in the above embodiment, and will not be elaborated herein.

[0109] Specifically, according to the embodiments disclosed by the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment disclosed by the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by a processing device 1001, it executes the above functions defined in the methods of the embodiments disclosed by the present invention.

[0110] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0111] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. An automotive generator performance detection method based on an intelligent sensing system, characterized in that, The method includes: Step S10: With the center of the core of the automotive generator stator winding as the center and a preset distance Δr, construct a dual-scale Hall magnetic field sensor array at positions with radii of R and R + Δr on the outer periphery of the stator winding core respectively, continuously and synchronously collect the magnetic field intensity data of each sensor in real time, and form a long-term dual-scale magnetic field feature matrix; Step S20: Use an improved empirical mode decomposition and singular spectrum analysis fusion method to extract features from the long-term dual-scale magnetic field feature matrix to obtain a performance trend feature matrix; Step S30: With the performance trend feature matrix obtained in Step S20 as the input, construct a performance evaluation model in the spatial frequency domain and output a continuous performance status scoring index; Step S40: Analyze the change of the performance status of the continuous performance status scoring index through a circular normal distribution and time series trend fusion method to obtain the performance status change trend; Step S50: According to the performance status change trend obtained in Step S40, combine reinforcement learning to construct a dynamic maintenance decision recommendation model and output a generator stator performance detection report.

2. The method for detecting the performance of an automotive generator based on an intelligent sensing system according to claim 1, characterized in that, In Step S10, 6 Hall magnetic field sensors are arranged at intervals of 60° in the inner circle with a radius of R, and 3 Hall magnetic field sensors are arranged at intervals of 120° in the outer circle with a radius of R + Δr, where Δr is 10 mm to 15 mm; the magnetic sensitive axis directions of all Hall magnetic field sensors point radially towards the center of the circle; the long-term dual-scale magnetic field feature matrix is expressed as: Among them, B R (·, t) is the magnetic field characteristic at the position with a radius of R at time t, and B R+Δr (·, t) is the magnetic field characteristic at the position with a radius of R + Δr at time t.

3. A method for detecting the performance of an automotive generator based on an intelligent sensing system according to claim 1, characterized in that, In Step S20, the steps of using an improved empirical mode decomposition and singular spectrum analysis fusion method to extract features from the long-term dual-scale magnetic field feature matrix to obtain a performance trend feature matrix specifically include: Step S201: Use an improved empirical mode decomposition method introducing a spatial period constraint condition to perform mode decomposition processing on the long-term dual-scale magnetic field feature matrix to obtain an effective intrinsic mode function component reflecting the long-term performance degradation trend; Step S202: Further perform singular value decomposition processing on the effective intrinsic mode function components by using the singular spectrum analysis fusion method to obtain a performance trend feature matrix reflecting the long-term degradation trend of the stator winding performance 4. The method for detecting the performance of an automotive generator based on an intelligent sensing system according to claim 1, characterized in that, In Step S30, the steps of constructing a performance evaluation model in the spatial frequency domain with the performance trend feature matrix obtained in Step S20 as the input specifically include: Construct a spatial domain attention mechanism layer: used to analyze the correlation of the performance trend feature matrix at different sensor spatial positions to obtain spatial weights; Construct a frequency domain attention mechanism layer: used to perform Fourier spectrum transformation on the performance trend feature matrix to automatically capture the frequency domain features of the magnetic field performance change and obtain frequency domain weights; Construct a Transformer fusion layer: used to form the final performance status evaluation feature by fusing the above spatial weights and frequency domain weights through an adaptive weighting method; Construct an output layer: used to output a continuous performance status scoring index PHI according to the performance status evaluation feature, where the range of PHI is from 0 to 1, which is used to represent the performance health status of the current stator winding, and a value close to 1 indicates a good performance status, and a value close to 0 indicates approaching performance degradation and failure.

5. A method for detecting the performance of an automotive generator based on an intelligent sensing system according to claim 1, characterized in that, In Step S30, the spatial frequency domain performance evaluation model is trained using historical measured data and simulated magnetic field performance data, including: pre-training the spatial frequency domain performance evaluation model using simulation data with a transfer learning strategy, and fine-tuning the model parameters using measured data.

6. The automotive generator performance detection method based on an intelligent sensing system according to claim 1, wherein In step S40, the step of analyzing the change in the performance state of the continuous performance state scoring index through the circular normal distribution and time series trend fusion method to obtain the performance state change trend specifically includes: Step S401: Construct a long-term performance state sequence based on the continuous performance state scoring index output in step S30, and use the circular normal distribution Von-Mises to perform probability density interpolation fitting on the long-term performance state sequence. The specific formula is: where θ is the angle value representing the mapping of the performance state scoring index to the circular space; μ is the mean parameter of the circular distribution; k is the concentration parameter of the circular distribution; I0(k) is the zero-order modified Bessel function; f(θ|μ,k) is the probability density interpolation fitting function; Step S402: Use the maximum likelihood estimation to solve the parameters μ and k of the circular normal distribution Von-Mises in step S401 to obtain an optimized probability density fitting model; Step S403: Use the optimized probability density fitting model obtained in step S402 to predict the trend of the performance state scoring index within a preset future time period through the time series prediction method to obtain the future performance change trend curve; Step S404: Calculate the performance attenuation degree and the corresponding remaining life confidence interval of the automotive generator stator winding within a specified future period according to the future performance change trend curve to obtain the performance state change trend.

7. The method for detecting the performance of an automotive generator based on an intelligent sensing system according to claim 1, characterized in that, In step S50, the step of constructing a dynamic maintenance decision recommendation model based on the performance state change trend obtained in step S40 and outputting a generator stator performance detection report specifically includes: Step S501: Obtain the historical maintenance operation record data and the historical maintenance operation effect data, use the performance state change trend as the input feature of the reinforcement learning state, and define the system state vector S in combination with the historical maintenance operation record data; Step S502: Define the maintenance action space A based on the historical maintenance operation record data, specifically including: A1: Do not perform maintenance for the time being and continue monitoring; A2: Local insulation cleaning and maintenance; A3: Replacement of local insulation components; A4: Local repair of the winding; A5: Overall replacement of the winding; Step S503: Construct a reward function R(S,A) according to the maintenance action space A in combination with the historical maintenance operation effect data; R(S,A) = w1ΔPHI recovery (A) - w2Cost(A) - w3Risk(S) where, ΔPHI recovery (A) is the expected performance score recovery value after performing action A; Cost(A) is the cost required to maintain action A; Risk(S) is the risk of the future performance state when maintenance is not performed; w1, w2, and w3 are adjustable weight parameters used to adjust the importance ratios of performance improvement, cost, and risk respectively; Step S504: Construct a dynamic maintenance decision recommendation model according to the reward function R(S,A). The dynamic maintenance decision recommendation model specifically includes: Input layer: Receive the system state vector; Hidden layer: Adopt a two-stream network structure, including: Stream A: Contains 2 fully connected layers, with the number of neurons in each layer being 64 and 32 respectively, and the activation function is ReLU, which is used to analyze the performance state trend; Stream B: Contains 2 fully connected layers, with the number of neurons in each layer being 64 and 32 respectively, and the activation function is ReLU, which is used to estimate the expected effect of each maintenance action; Fusion layer: Concatenate the feature vectors output by Stream A and Stream B to form a joint state-action evaluation vector; Output layer: The output layer is a fully connected layer, which is used to calculate the action value Q value of each maintenance action, generate a generator stator performance detection report and output it.

8. An automotive generator performance detection system based on an intelligent sensing system, which is applied to an automotive generator performance detection method based on an intelligent sensing system according to any one of claims 1-7, characterized in that, The vehicle generator performance detection system based on the intelligent sensing system includes: A magnetic field data acquisition module, which is used to construct a dual-scale Hall magnetic field sensor array at positions with radii of R and R+Δr respectively on the outer circumference of the stator winding core with the center of the stator winding core of the vehicle generator as the center of the circle, continuously and real-time synchronously collect the magnetic field intensity data of each sensor, and form a long-term dual-scale magnetic field feature matrix; A performance feature extraction module, which is used to extract features from the long-term dual-scale magnetic field feature matrix by using an improved empirical mode decomposition and singular spectrum analysis fusion method to obtain a performance trend feature matrix; A state scoring calculation module, which is used to construct a spatial frequency domain performance evaluation model with the performance trend feature matrix obtained in step S20 as the input and output a continuous performance state scoring index; A state change analysis module, which is used to analyze the performance state change of the continuous performance state scoring index by using a circular normal distribution and time series trend fusion method to obtain a performance state change trend; A detection report generation module, which is used to construct a dynamic maintenance decision recommendation model by combining reinforcement learning according to the performance state change trend obtained in step S40 and output a generator stator performance detection report.

9. An automotive generator performance detection device based on an intelligent sensing system, characterized in that, The vehicle generator performance detection device based on the intelligent sensing system includes: a memory, a processor, and a vehicle generator performance detection program based on the intelligent sensing system stored on the memory and executable on the processor. When the vehicle generator performance detection program based on the intelligent sensing system is executed by the processor, it implements the vehicle generator performance detection method according to any one of claims 1 to 7.

10. A computer program product, characterized in that The computer program product includes a vehicle generator performance detection program based on the intelligent sensing system. When the vehicle generator performance detection program based on the intelligent sensing system is executed by the processor, it implements the vehicle generator performance detection method according to any one of claims 1 to 7.

Images