AMT bearing operation state detection system and method under active and passive switching working condition

Abstract

The invention relates to the technical field of automobile automatic transmission, and discloses an AMT bearing operation state detection system and method under an active and passive switching working condition, and the method comprises the steps: collecting real-time data through a vibration acceleration sensor, a current sensor, a temperature sensor and a rotating speed encoder, carrying out the preprocessing of envelope demodulation, Kalman filtering and the like, and carrying out the detection of the operation state of an AMT bearing; and generating a time-frequency characteristic matrix by using variational mode decomposition and Hilbert transform. And inputting the matrix into a probabilistic neural network model adopting a sliding time window mechanism, and outputting a bearing health state probability value. A multi-parameter state evaluation model optimized by a quantum genetic algorithm is constructed, an optimal feature combination is obtained, and early warning levels and maintenance suggestions are output through a belief rule base inference device in combination with a hierarchical diagnosis control model (including an acquisition layer, an analysis layer and an execution layer). The system is further provided with a signal verification module to ensure data reliability. According to the invention, multi-source data fusion and dynamic adaptive diagnosis are realized, and the state detection precision and real-time performance of the AMT bearing under complex working conditions are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic transmission technology for automobiles, and in particular to a system and method for detecting the operating status of an AMT bearing under active and passive switching conditions. Background Art

[0002] As the automotive industry evolves toward intelligent and automated driving, automatic mechanical transmissions (AMTs) are gaining widespread adoption due to their high transmission efficiency and low cost. However, monitoring the operating status of AMT bearings under complex operating conditions has long been a challenge. During active and passive switching operations (such as starting, shifting, and braking), the load, speed, and impact loads on the bearings fluctuate dramatically, making it difficult for traditional detection methods to effectively detect early fault characteristics, leading to transmission failure and even accidents.

[0003] In existing technologies, single-sensor detection methods (such as those relying solely on vibration signals) are susceptible to interference from operating conditions and cannot fully reflect the operating status of the bearing. Diagnostic strategies based on fixed thresholds lack the ability to adapt to real-time operating conditions, and false positives or omissions are common during active-passive switching. Traditional feature extraction algorithms (such as Fourier transforms) have limitations when processing non-stationary signals, making it difficult to accurately characterize the changes in time-frequency domain characteristics caused by bearing faults. In addition, the high feature redundancy and low computational efficiency in the multi-parameter fusion process make it difficult for existing systems to meet the real-time and reliability requirements of AMT transmissions.

[0004] With the development of intelligent algorithms, some research has attempted to apply neural networks to bearing fault diagnosis. However, traditional neural network models suffer from low parameter update efficiency under dynamic conditions and a fixed hidden layer node activation threshold, making them unable to adapt to the real-time changes in AMT bearing loads. Genetic optimization algorithms suffer from premature convergence during feature selection, making it difficult to achieve global optimality in multi-parameter spaces. During the inference process of a confidence rule base, fixed rule weights result in a lack of dynamic adjustment capabilities for diagnostic strategies, making it impossible to fully utilize historical diagnostic data to optimize the diagnostic process.

[0005] At the data preprocessing level, existing methods fail to adequately extract modulation features from vibration signals, current signals are susceptible to electromagnetic interference, and noise suppression in temperature and speed data is ineffective, resulting in low feature quality for the input model. In a hierarchical diagnostic system, multi-level diagnostic strategies lack connectivity and a mechanism for predicting the evolution of bearing health, making it difficult to trigger effective maintenance measures in the early stages of a fault.

[0006] The existing technology in AMT bearing operating status detection has problems such as insufficient multi-source data fusion, insufficient feature extraction accuracy, weak model adaptability, and poor real-time diagnosis. There is an urgent need for an intelligent detection system and method that can adapt to active and passive switching conditions, integrate multi-sensor data, and have dynamic optimization capabilities to improve the reliability and service life of AMT gearboxes. Summary of the Invention

[0007] The purpose of the present invention is to provide an AMT bearing operating status detection system and method under active and passive switching conditions to solve the problems raised in the above background technology.

[0008] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a method for detecting the operating status of an AMT bearing under active-passive switching conditions, the method comprising: collecting real-time operating data of a gearbox bearing through a multi-channel sensing module, the multi-channel sensing module comprising a vibration acceleration sensor, a current sensor, a temperature sensor and a speed encoder; extracting intrinsic mode components of the real-time operating data based on a variational mode decomposition algorithm, and generating a time-frequency feature matrix through a Hilbert transform; inputting the time-frequency feature matrix into a pre-trained probabilistic neural network model, the probabilistic neural network model adopts a sliding time window mechanism, adjusts the hidden layer node activation threshold based on the real-time operating condition data, and outputs a bearing health status probability value; and outputs a bearing health status probability value according to the bearing health status probability. A multi-parameter state assessment model is constructed based on the time-frequency feature value, wherein the multi-parameter state assessment model takes the highest fault sensitivity and the lowest feature redundancy as the optimization goals, and uses a quantum genetic optimization algorithm to perform a global search on the feature subset, wherein the quantum genetic optimization algorithm introduces an adaptive rotation angle and a dynamic crossover probability; based on the multi-parameter state assessment model, optimal feature combination data is output; according to the optimal feature combination data, a hierarchical diagnosis and control model is established, wherein the hierarchical diagnosis and control model includes an acquisition layer, an analysis layer and an execution layer, wherein the acquisition layer updates the operating mode based on the time-frequency feature matrix, the analysis layer generates a multi-level diagnosis strategy based on the bearing health state probability value, and the execution layer maps the diagnostic instructions into warning levels and maintenance recommendation signals through a confidence rule base reasoner.

[0009] Preferably, the probabilistic neural network model adopts a sliding time window mechanism including: constructing a probabilistic network structure, the probabilistic network structure includes a vibration feature node, a current feature node and a health status node, the vibration feature node and the current feature node are parent nodes, and the health status node is a child node; updating the probability density function of the vibration feature node based on real-time collected data, performing frequency band energy analysis on the current feature node through a wavelet packet decomposition algorithm to obtain the posterior probability of the characteristic frequency band energy distribution; introducing a dynamic threshold adjustment mechanism to dynamically adjust the inference step of the probabilistic network according to the change of the bearing load, and the time window width is negatively correlated with the bearing speed; using the expectation maximization algorithm to perform online parameter update on the probabilistic neural network, and optimizing the hidden layer node weight parameters by maximizing the likelihood function.

[0010] Preferably, the quantum genetic optimization algorithm introduces adaptive rotation angle and dynamic crossover probability, including: constructing a genetic optimization objective function, the objective function includes a feature sensitivity term and a feature independence term, the feature sensitivity term is calculated by the inter-class distance of the fault sample, and the feature independence term is calculated by the correlation coefficient between the feature vectors; designing an adaptive rotation angle update rule, the rotation angle changes in an exponential decay curve with the increase of evolutionary generations, maintaining a large search range in the initial stage, and converging to the optimal solution neighborhood in the later stage; designing a dynamic crossover probability adjustment mechanism, the crossover probability is dynamically adjusted according to the population fitness variance, and the crossover operation frequency is increased when the population variance is higher than the threshold, and vice versa.

[0011] Preferably, the confidence rule base reasoner maps the diagnostic instructions into warning levels and maintenance recommendation signals, including: constructing an input activation module, the input including the probability value of the bearing health status, the temperature change gradient and the bearing operating time, and using the Gaussian membership function to assign confidence to the input variables; constructing a rule base, the rule base containing confidence rules; designing an evidence fusion module, using DS evidence theory to fuse multi-source rule outputs into a comprehensive confidence distribution, the output including the warning level code and the maintenance recommendation code; introducing a rule weight correction mechanism, and dynamically updating the rule confidence coefficient according to the historical diagnosis accuracy and maintenance records.

[0012] Preferably, the rule construction of the confidence rule base reasoner includes: establishing a fault mode knowledge base, wherein the knowledge base contains vibration spectrum feature templates of three typical bearing faults: pitting, cracking, and wear.

[0013] Preferably, the data preprocessing of the multi-channel sensing module includes: using an envelope demodulation algorithm to extract fault characteristic frequency components from vibration acceleration data; using fast Fourier transform to generate a spectrum diagram for current sensor data, and suppressing background noise through adaptive threshold filtering; using an exponentially weighted moving average algorithm to eliminate measurement fluctuations for temperature sensor data; using a Kalman filter algorithm to fuse angular velocity and angular acceleration data for speed encoder data, and outputting the instantaneous speed and speed fluctuation rate of the bearing.

[0014] Preferably, the analysis layer of the hierarchical diagnostic control model generates a multi-level diagnostic strategy including: dividing the health level into three levels: normal, warning, and fault, the normal level triggers periodic monitoring, the warning level triggers feature tracking monitoring, and the fault level triggers the shutdown protection protocol; constructing a state transfer matrix, the matrix performs state jumps according to continuous health level changes, and if the health level drops for three consecutive sampling cycles, an upgraded diagnostic process is triggered; introducing a trend prediction mechanism, and starting the active test mode when the change rate of the bearing health status probability value exceeds a preset gradient threshold.

[0015] Preferably, the constraints of the multi-parameter state assessment model include: feature dimension constraint, which limits the maximum dimension of the feature vector to no more than the upper threshold of computing resource allocation; operating condition coverage constraint, which verifies the operating condition adaptability of the feature combination through typical operating condition simulation tests; and real-time constraint, which limits the maximum delay of a single diagnostic process to no more than 20% of the control period.

[0016] Preferably, the method further comprises: constructing a signal verification module, detecting signal anomalies through multi-sensor data correlation analysis, and initiating data reconstruction if the correlation coefficient between a certain channel data and the associated channel is lower than a set threshold.

[0017] Preferably, the present invention also includes an AMT bearing operating status detection system under active-passive switching conditions, the system comprising: a multi-channel sensing module: for collecting real-time operating data of the gearbox bearing, comprising a vibration acceleration sensor, a current sensor, a temperature sensor and a speed encoder; a feature extraction module: extracting the intrinsic mode components of the real-time operating data collected by the multi-channel sensing module based on the variational mode decomposition algorithm, and generating a time-frequency feature matrix through Hilbert transform; a state evaluation module: inputting the time-frequency feature matrix into a pre-trained probabilistic neural network model, which adopts a sliding time window mechanism to adjust the hidden layer node activation threshold based on the real-time operating condition data. value, and output the bearing health status probability value; optimization decision module: constructs a multi-parameter state assessment model based on the bearing health status probability value, takes the highest fault sensitivity and the lowest feature redundancy as the optimization goal, adopts the quantum genetic optimization algorithm that introduces adaptive rotation angle and dynamic crossover probability to perform a global search on the feature subset, and outputs the optimal feature combination data; diagnosis execution module: is established based on the optimal feature combination data, including the acquisition layer, analysis layer and execution layer, among which the acquisition layer updates the operating condition mode based on the time-frequency feature matrix, the analysis layer generates a multi-level diagnosis strategy based on the bearing health status probability value, and the execution layer maps the diagnostic instructions to the warning level and maintenance recommendation signal through the confidence rule base reasoner.

[0018] Compared with existing technologies, the present invention offers the following advantages: During data acquisition and preprocessing, a multi-channel sensing module integrates vibration acceleration, current, temperature, and speed encoders, enabling comprehensive awareness of the bearing's operating status. Based on the data characteristics of different sensor types, algorithms such as envelope demodulation, fast Fourier transform, exponentially weighted moving average, and Kalman filtering are used for preprocessing, effectively suppressing noise and extracting key features. For example, the envelope demodulation algorithm for vibration signals enhances fault-specific frequency components, while adaptive threshold filtering for current signals effectively suppresses electromagnetic interference, ensuring that the input model's feature data possesses a high signal-to-noise ratio and physical characterization capabilities.

[0019] During the feature extraction and state assessment phase, the combination of variational mode decomposition (VMD) and the Hilbert transform accurately extracts intrinsic mode components from nonstationary operating data and generates a time-frequency feature matrix, overcoming the shortcomings of traditional Fourier transforms in processing nonlinear signals. The probabilistic neural network model introduces a sliding time window mechanism, dynamically adjusting the hidden layer node activation threshold and inference step size to enable the model to adaptively update parameters based on real-time operating conditions (such as bearing speed and load changes). Dynamic updates of the probability density functions of vibration and current feature nodes and wavelet packet energy analysis enhance the model's sensitivity to feature changes under active-passive switching conditions. The outputted bearing health probability value provides a quantitative basis for subsequent evaluation.

[0020] The multi-parameter state assessment model optimizes for maximum fault sensitivity and minimum feature redundancy using a quantum genetic optimization algorithm that incorporates an adaptive rotation angle and dynamic crossover probability. The adaptive rotation angle decays exponentially with the number of evolutionary generations, balancing global search capability in the initial phase with local convergence accuracy in the later stages. The dynamic crossover probability is adjusted based on the population fitness variance, avoiding the premature convergence problem of traditional genetic algorithms and ensuring the optimal feature combination is found in the high-dimensional feature space. This model simultaneously meets the constraints of feature dimensionality, operating condition coverage, and real-time performance, improving diagnostic accuracy while reducing computational complexity, meeting the stringent real-time requirements of AMT transmissions.

[0021] The hierarchical diagnostic control model constructs a multi-level diagnostic system through the collaborative work of the acquisition layer, analysis layer, and execution layer. The analysis layer divides health levels into three levels: normal, warning, and fault. Combining a state transition matrix and trend prediction mechanism, it achieves dynamic policy switching from periodic monitoring to shutdown protection. Continuous declines in health levels trigger an upgraded diagnostic process, and an active test mode is initiated when the probability value change rate exceeds a threshold. This forward-looking design provides timely warnings at the early stages of a fault. The confidence rule base reasoner at the execution layer integrates multi-source inputs (health state probability values, temperature gradients, and operating time), achieves evidence fusion through DS evidence theory, and dynamically adjusts rule weights based on historical data, significantly improving the reliability of diagnostic conclusions and the pertinence of maintenance recommendations.

[0022] The introduction of the signal verification module is another innovation of the present invention. It detects signal anomalies through multi-sensor data correlation analysis. When the correlation coefficient between a channel data and the associated channel data is lower than the threshold, data reconstruction is initiated to ensure that the data source of the input system is reliable and avoid misdiagnosis due to a single sensor failure.

[0023] This invention combines multi-source data fusion, intelligent algorithm optimization, dynamic threshold adjustment, and a hierarchical diagnostic strategy to construct a comprehensive detection system suitable for active and passive switching conditions in AMT bearings. This solution not only addresses the adaptability challenges of traditional methods under non-stationary conditions but also improves diagnostic accuracy through feature optimization and model adaptation. This provides an advanced technical approach for ensuring the reliability of AMT transmissions, with significant engineering application value and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a working principle diagram of the AMT bearing operating status detection method under active and passive switching conditions described in the present invention;

[0025] Figure 2 This is a diagram showing the working principle of the sliding time window mechanism of the probabilistic neural network model;

[0026] Figure 3 Design diagram of the confidence rule base reasoner. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] See also Figure 1-Figure 3 The present invention relates to a method for detecting the operating status of an AMT bearing under active-passive switching conditions, which achieves accurate detection and diagnosis of the bearing operating status through multi-module collaboration. Specifically, it includes the following steps:

[0029] Real-time operating data on the gearbox bearings is collected through a multi-channel sensing module, which includes a vibration accelerometer, current sensor, temperature sensor, and speed encoder. These sensors are located in the gearbox bearing housing, motor input circuit, bearing cavity, and drive shaft end. The vibration accelerometer is mounted in three directions to obtain complete vibration information. The current sensor uses a Hall effect element to monitor motor current fluctuations in real time. The temperature sensor uses a PT100 thermal resistor embedded in the bearing lubrication cavity. The speed encoder is rigidly connected to the drive shaft via an incremental code disk.

[0030] The variational mode decomposition (VMD) algorithm is used to extract intrinsic mode components from real-time operating data, and a time-frequency feature matrix is ​​generated through a Hilbert transform. In specific implementation, the VMD algorithm parameters are set to: mode number K = 6, quadratic penalty factor α = 2000, and adaptive decomposition of each modal component is achieved through frequency domain masking and quadratic programming. The time-frequency feature matrix is ​​input into a pretrained probabilistic neural network model. This model uses a sliding time window mechanism to adjust the hidden layer node activation threshold based on real-time operating data and output a bearing health status probability value. The time window width is initially set to 1024 sampling points. The bearing speed is calculated in real time using a speed encoder, and the window width is dynamically adjusted based on a negative correlation (window width = 1024 × reference speed / real-time speed).

[0031] A multi-parameter condition assessment model was constructed based on the bearing health status probability values. With the optimization goals of maximizing fault sensitivity and minimizing feature redundancy, a quantum genetic optimization algorithm was used to perform a global search for feature subsets. During the optimization process, the initial population size was set to 50, and the number of evolutionary generations was 100. Adaptive rotation angles and dynamic crossover probability mechanisms were introduced to improve search efficiency. The final output was the optimal feature combination data, including parameters such as vibration energy entropy, current harmonic distortion, and temperature gradient.

[0032] A hierarchical diagnostic control model was established based on the optimal feature combination data. This model consists of an acquisition layer, an analysis layer, and an execution layer. The acquisition layer identifies operating modes (active / passive switching) based on a time-frequency feature matrix. The analysis layer generates a three-level diagnostic strategy (normal, warning, and fault) based on health state probabilities. The execution layer uses a confidence rule base reasoner to map diagnostic instructions into warning levels (Level I, Level II, and Level III) and maintenance recommendation signals (continue operation, regular inspection, and immediate shutdown).

[0033] The present invention will be further described below in conjunction with Examples 1 to 5:

[0034] Example 1: In implementing the sliding time window mechanism of a probabilistic neural network model, a probabilistic network structure is first constructed, comprising a vibration feature node, a current feature node, and a health status node. The vibration feature node and the current feature node serve as parent nodes, and the health status node serves as a child node. The vibration feature node receives signal data collected by a vibration acceleration sensor and specifically extracts statistical features such as the root mean square value and kurtosis value of the vibration signal as input parameters. These parameters reflect the energy intensity and impact characteristics of the vibration signal. The current feature node receives current signal data collected by the current sensor and extracts features such as the fundamental component and the energy of each harmonic as input parameters. The fundamental component reflects the basic operating state of the current, while the energy of each harmonic reflects electromagnetic anomalies during motor operation. Based on the parameters input by the parent node, the health status node outputs a health probability value between 0 and 1 through internal network calculation and reasoning. This value represents the likelihood that the bearing is currently in a healthy state, with a larger value indicating a better bearing health status.

[0035] Based on the real-time data collected, the probability density function of the vibration feature nodes needs to be updated. This is accomplished by using kernel density estimation to divide the vibration data into several equally spaced intervals, for example, 50 intervals, according to a specific rule. Then, for each sample point within each interval, the weighted sum of the Gaussian kernel function is calculated. This generates a continuous probability density curve, which more accurately describes the probability distribution of the vibration features, enabling the probabilistic neural network model to adapt to changes in the vibration signal in real time.

[0036] For current characteristic nodes, a wavelet packet decomposition algorithm is used to perform frequency band energy analysis. First, the number of wavelet packet decomposition levels is determined. For example, setting the decomposition level to three yields eight characteristic frequency bands. Next, the energy distribution of each characteristic frequency band is calculated, and the posterior probability corresponding to each frequency band energy is further calculated using the Bayesian formula. This decomposes the current signal into different frequency bands for analysis, extracting frequency band energy features with greater diagnostic value and providing richer information for determining bearing health status.

[0037] To enable the probabilistic neural network model to better adapt to changes in bearing load, a dynamic threshold adjustment mechanism is introduced. Bearing load changes can be characterized by the root mean square value of the current signal. When the load change rate exceeds a set threshold (e.g., 5%), it indicates a significant change in the bearing's operating condition. At this point, the probabilistic network's inference step size needs to be adjusted from a longer time (e.g., 10ms) to a shorter time (e.g., 5ms) to enable more timely analysis and judgment of the bearing's condition.

[0038] During the model's operation, the expectation-maximization algorithm is used to perform online parameter updates on the probabilistic neural network. This algorithm iteratively calculates the maximum likelihood estimates of the hidden layer node weight parameters. The specific process is as follows: first, the expectation of the data is calculated based on the current parameter estimates, and then the parameters are updated by maximizing this expectation. This process is repeated continuously, allowing the model parameters to be continuously optimized based on real-time data, thereby improving the model's accuracy and adaptability.

[0039] The specific implementation of the sliding time window mechanism also involves dynamic adjustment of the time window width. The time window width is negatively correlated with the bearing speed: higher bearing speeds result in narrower time window widths, while lower speeds result in wider time window widths. In practical applications, the speed encoder first acquires real-time bearing speed data. The current time window width is then calculated based on a preset negative correlation formula (e.g., the time window width equals the initial width multiplied by the reference speed, divided by the real-time speed. The initial width can be set to include a certain number of sampling points, such as 1024). This dynamic adjustment ensures that the time window captures data appropriately under different speed conditions, ensuring both data timeliness and fully capturing the changing characteristics of the bearing's operating state.

[0040] When constructing a probabilistic network structure, it's important to clearly define the connections between nodes and the calculation methods. Parent nodes (vibration and current feature nodes) and child nodes (health status nodes) are connected by weights, which reflect the impact of the parent node's characteristics on the bearing's health. During model training, these weights are learned and optimized using extensive sample data, enabling the model to accurately output reasonable health status probabilities based on the parent node's input features.

[0041] During data processing, the raw data collected by vibration accelerometers and current sensors requires preprocessing to remove noise and interference and improve data quality. For example, vibration data can be de-averaged and filtered to remove DC components and high-frequency noise; current data can be normalized to make current signals of different amplitudes comparable. This preprocessed data is then input into a probabilistic network structure for feature extraction and state assessment.

[0042] Furthermore, to ensure the stability and reliability of the sliding time window mechanism, the time window's moving step size must be appropriately set. The moving step size is typically smaller than the time window width. This ensures a certain amount of data overlap and prevents loss of important information due to data truncation. For example, if the time window width is 1024 sampling points, the moving step size can be set to 512 sampling points. This way, each time the time window moves, half of the data is new and half is old, ensuring both real-time performance and data continuity.

[0043] During the model's online operation, the bearing's operating status needs to be monitored in real time, and timely warnings or appropriate measures should be issued based on changes in the health probability value. For example, when the health probability value falls below a certain threshold, it indicates that the bearing may be abnormal and requires further diagnosis and analysis. When the probability value continues to decrease, it may indicate the occurrence of a bearing failure, and maintenance personnel must be notified for inspection and repair.

[0044] Example 2: Introducing adaptive rotation angle and dynamic crossover probability mechanisms into the quantum genetic optimization algorithm aims to improve the efficiency and accuracy of the global search for feature subsets, thereby constructing a multi-parameter state assessment model that meets the goals of maximum fault sensitivity and minimum feature redundancy. Specific implementation methods are as follows:

[0045] First, a genetic optimization objective function is constructed, consisting of two components: a feature sensitivity term and a feature independence term. The feature sensitivity term is calculated using the inter-class distance of fault samples. This inter-class distance measures the degree of separation between samples of different fault categories in the feature space; a larger distance indicates a higher sensitivity of the feature to fault classification. The feature independence term is calculated using the correlation coefficient between feature vectors. This coefficient reflects the degree of linear correlation between features; a smaller coefficient indicates lower redundancy between features. By combining these two metrics, an optimization objective is formed for the feature subset, ensuring that the selected features can effectively distinguish fault categories while minimizing information overlap between features.

[0046] The design of the adaptive rotation angle update rule needs to consider the needs of different stages of the evolutionary process. In the early stages of evolution, the population is not yet close to the optimal solution and requires a larger search range to explore a wider feature space. Therefore, a larger rotation angle is set in the initial stage so that the algorithm can conduct a global search in a larger range and avoid falling into a local optimal solution too early. As the number of evolutionary generations increases, the population gradually converges to the neighborhood of the optimal solution. At this time, the rotation angle is reduced to make the search step size thinner, thereby performing a more refined local search near the optimal solution and improving the optimization accuracy. The change in the rotation angle follows a specific curve law, such as an exponential decay curve as the number of evolutionary generations increases. In this way, the search range can be adaptively adjusted from global to local, balancing the breadth and depth of the search.

[0047] The core of the dynamic crossover probability adjustment mechanism is to dynamically adjust the frequency of crossover and mutation operations based on the population fitness variance. The population fitness variance reflects the degree of variability between individuals in the population. When the variance is above a set threshold, it indicates a high diversity of individuals in the population and the presence of many different feature combinations. In this case, the frequency of crossover operations is increased, combining the best features of different individuals through crossover operations to further explore new feature spaces and enhance global search capabilities. When the variance is below the threshold, it indicates that individuals in the population are becoming more similar, diversity is decreasing, and the population is prone to falling into local optima. In this case, the frequency of mutation operations is increased, introducing new feature combinations through mutation operations, breaking the stagnant state of the population and improving local search accuracy. This dynamic adjustment method enables the algorithm to automatically switch search strategies based on the actual state of the population, improving optimization efficiency.

[0048] During implementation, it's necessary to first determine the basic parameters of the genetic algorithm, such as the initial population size and the number of generations. The initial population size can be set to a certain number of individuals, each representing a subset of features. For example, the initial population size can be set to 50 individuals, with each individual consisting of a binary code indicating whether a feature is selected (e.g., 1 for selected features, 0 for unselected features). The number of generations should be set to a sufficiently large value to ensure the algorithm has sufficient time to search and optimize. For example, the number of generations can be set to 100.

[0049] During each generation of evolution, the fitness of each individual is first calculated. This is the combined score of fault sensitivity and feature independence for the feature subset corresponding to that individual, calculated based on the genetic optimization objective function. Individuals with higher fitness values ​​are associated with feature subsets that are closer to the optimization target. Then, a selection operation is performed based on these fitness values. Using methods such as roulette wheel selection and tournament selection, individuals with higher fitness are selected from the population as parents, providing the basis for subsequent crossover and mutation operations.

[0050] During a crossover operation, a dynamic crossover probability adjustment mechanism determines whether to perform a crossover on the selected parent individuals. A crossover exchange partially encodes two parent individuals to generate new offspring individuals, thereby generating new feature combinations. For example, for two binary-coded parent individuals, single-point or multi-point crossover methods can be used to exchange the encoding segments of both parents at randomly selected crossover points to generate two new offspring individuals. Because the crossover probability is dynamically adjusted based on the variance of the population's fitness, when the variance is high, the crossover probability is higher, resulting in more parent individuals undergoing crossover operations and generating more new feature combinations. When the variance is low, the crossover probability is lower, reducing the frequency of crossover operations and avoiding over-exploration that can lead to decreased search efficiency.

[0051] Mutation randomly changes the encoding of an individual, for example, changing a binary 0 to a 1 or vice versa, thereby introducing new features or removing selected features. The probability of mutation is typically set low to avoid destroying optimal feature combinations. However, when the variance of a population's fitness is low, increasing the frequency of mutation can effectively improve population diversity. For example, when the variance is below a threshold, the mutation probability can be increased from an initial low value (e.g., 0.01) to an appropriate value (e.g., 0.05) to increase the probability of generating new feature combinations.

[0052] Throughout the evolutionary process, the adaptive rotation angle mechanism continuously adjusts the rotation angle to alter the individual's search direction and step size in the feature space. For example, in the early stages of evolution, the rotation angle is large, and the individual's movement step size in the feature space is large, allowing for rapid exploration of different feature regions. As the number of evolutionary generations increases, the rotation angle gradually decreases, and the individual's movement step size becomes finer, gradually focusing on feature regions near the optimal solution and making fine adjustments. This mechanism enables the algorithm to automatically adjust its search strategy at different stages of evolution, improving search efficiency and optimization quality.

[0053] In addition, attention should be paid to the impact of conditions such as feature dimension constraints, working condition coverage constraints, and real-time constraints on the optimization process. Feature dimension constraints limit the maximum dimension of the feature vector, for example, not exceeding the upper threshold of computing resource allocation. During the optimization process, it is necessary to ensure that the number of selected features does not exceed this threshold to avoid the computational complexity exceeding the system's tolerance due to excessive feature dimensions. Working condition coverage constraints require that the working condition adaptability of feature combinations be verified through typical working condition simulation tests. Therefore, during the optimization process, the performance of features under different working conditions must be considered to ensure that the selected features can effectively reflect the operating status of the bearing under various typical working conditions. Real-time constraints limit the maximum delay of a single diagnostic process. Therefore, when designing the algorithm, the computational efficiency of the algorithm must be optimized to ensure that each step of the evolution process can be completed within a reasonable time to meet the real-time requirements of the system.

[0054] By introducing an adaptive rotation angle and dynamic crossover probability mechanism, the quantum genetic optimization algorithm can automatically adjust the search strategy based on the evolutionary stage and population status during the global search of feature subsets, effectively balancing global and local search capabilities and improving the efficiency and quality of feature optimization. This mechanism enables the multi-parameter state assessment model to quickly screen the optimal feature combination data, providing reliable feature input for subsequent hierarchical diagnostic control, thereby improving the accuracy and reliability of the entire AMT bearing operation status detection system. In practical applications, by reasonably setting algorithm parameters and constraints, this mechanism can effectively adapt to the complex requirements of bearing operation status detection under active-passive switching conditions, providing strong support for bearing fault diagnosis and maintenance.

[0055] Example 3: In implementing the confidence rule base reasoner to map diagnostic instructions into warning levels and maintenance recommendation signals, an input activation module must first be constructed. Its input parameters include the bearing health probability value, temperature gradient, and bearing operating time. The bearing health probability value, output by the probabilistic neural network model and ranging from 0 to 1, represents the likelihood that the bearing is healthy. The temperature gradient, calculated from temperature sensor data and expressed in °C / min, reflects the rate of change of bearing temperature over time. The bearing operating time, expressed in hours, is calculated by accumulating the operating time of the bearing since startup and is used to assess the impact of the bearing's service life on its condition.

[0056] For the above input parameters, the Gaussian membership function is used to assign confidence to them, so as to convert the precise input data into fuzzy confidence values, which is convenient for the rule base to reason. The general form of the Gaussian membership function is: in, is the specific value of the input variable, is the central value of the corresponding fuzzy subset, representing the typical eigenvalue of the fuzzy subset; is the width parameter, which determines the “fatness” of the membership function curve and reflects the range of the fuzzy subset. For example, for the probability value of the bearing health status, three fuzzy subsets of “normal”, “warning” and “fault” can be defined. The center value of the “normal” subset is Set to 0.8, width Set to 0.1, it means that when the health probability value is close to 0.8, the confidence level of the “normal” subset is the highest, and as the value deviates from 0.8, the confidence level gradually decreases; the central value of the “fault” subset Set to 0.3, width Set to 0.1 to describe a poor health state. For the temperature gradient and bearing operating time, the corresponding fuzzy subsets, center values, and width parameters should also be set based on engineering experience and historical data. For example, the temperature gradient can be divided into three subsets: "low," "medium," and "high," and the bearing operating time can be divided into three subsets: "short," "medium," and "long," each corresponding to a different center value and width parameter.

[0057] Next, a rule base is constructed. This base contains multiple confidence rules. Each rule describes the mapping between input variables and output results in the form of an "IF condition, THEN conclusion" formula. For example, a typical rule might be formulated as: "IF the bearing health probability value is 'normal' AND the temperature gradient is 'low' AND the bearing operating time is 'short' THEN the warning level is Level I, and the confidence level is 0.8." The rules in the rule base are constructed based on a fault mode knowledge base and domain expert experience. The fault mode knowledge base contains vibration spectrum feature templates for three typical bearing faults: pitting, cracking, and wear. For example, pitting faults correspond to a characteristic frequency of 1.5 times the rotational frequency, crack faults to 2 times the rotational frequency, and wear faults to 0.8 times the rotational frequency. These feature templates serve as important basis for rule conditions. The rule base initially contains a certain number of rules, for example, 96 rules, covering diagnostic scenarios under different input combinations. The confidence coefficient for each rule (such as 0.8 in the above example) is initially set based on expert experience and typically ranges from 0.7 to 0.9, indicating the confidence level of the rule.

[0058] When designing the evidence fusion module, the DS evidence theory was used to fuse the outputs of multiple rules into a comprehensive confidence distribution. DS evidence theory is an inference method for processing uncertain information. It fuses the outputs of multiple rules by calculating basic probability assignment functions, trust functions, and likelihood functions. The specific process is as follows: First, for each rule's output (such as warning level and confidence), a corresponding basic probability assignment (BPA) is generated, representing the rule's support for a given warning level. Then, the Dempster synthesis rule is used to synthesize multiple basic probability assignments to obtain a comprehensive basic probability assignment. Finally, the trust function and likelihood function are calculated based on the comprehensive basic probability assignment to generate a comprehensive confidence distribution, which reflects the combined support of all rules for each warning level. The evidence fusion module effectively integrates information from multiple rules, improving the reliability and accuracy of diagnostic results.

[0059] To enable the confidence rule base reasoner to adapt to changes in actual operating conditions, a rule weight correction mechanism is introduced to dynamically update the rule confidence coefficient based on historical diagnostic accuracy and maintenance records. The specific correction method is: when a rule's diagnostic result is verified as correct through maintenance records, the confidence coefficient of the rule is increased; if the diagnostic result is incorrect, the confidence coefficient is reduced. The mathematical expression for the correction is: in, is the updated rule confidence coefficient, is the confidence coefficient before updating, is the learning rate, which is used to control the magnitude of the correction and usually ranges from 0 to 1 (e.g. ); The historical diagnostic accuracy is calculated by counting the ratio of the number of correct diagnoses in historical cases to the total number of diagnoses for the rule, with a value ranging from 0 to 1. Through this dynamic correction mechanism, the rule base can continuously learn from actual diagnostic experience and gradually improve the accuracy and applicability of the rules.

[0060] During implementation, parameter preprocessing of the input activation module is crucial. For the bearing health status probability value, the output of the probabilistic neural network model must be directly received and range-checked to ensure it is between 0 and 1. For the temperature gradient, a differential calculation is performed on the real-time data from the temperature sensor. For example, the temperature change per unit time is calculated using the temperature values ​​of two adjacent sampling points, and then smoothed to remove noise interference. For the bearing operating time, a cumulative calculation is performed using the system clock and the bearing start and stop signals. The current time is saved when the system is powered off to ensure data continuity.

[0061] The construction of a rule base must adhere to the principles of completeness and consistency. Completeness requires that the rule base cover all possible input combination scenarios to avoid diagnostic blind spots. Consistency requires that there are no conflicting conclusions between rules. For example, the same input combination cannot correspond to different warning levels at the same time. To meet these requirements, when building the rule base, an orthogonal experimental design method can be used to generate input combinations. Domain experts can then evaluate the diagnostic conclusions corresponding to each combination to ensure the comprehensiveness and rationality of the rules.

[0062] The computational process of the evidence fusion module must consider computational efficiency and numerical stability. Because DS evidence theory exhibits high computational complexity when synthesizing multiple pieces of evidence, a recursive synthesis approach can be used to gradually merge basic probability distributions to reduce computational complexity. Furthermore, the input basic probability distributions must be normalized to avoid bias in the synthesis results due to numerical errors. Furthermore, a confidence threshold can be set. When the confidence level of a warning level in the integrated confidence distribution exceeds the threshold, that level is directly output as the diagnosis result, improving diagnostic efficiency.

[0063] The vibration spectrum feature templates of the fault mode knowledge base are established using extensive experimental data and failure cases. For example, for pitting faults, pitting defects are artificially created on the bearing. Vibration signals are collected at different speeds and loads, and spectrum analysis is performed to determine the characteristic frequency as 1.5 times the rotational frequency. The manifestation of this characteristic under different operating conditions is recorded. For crack and wear faults, similar methods are used to obtain characteristic frequency and spectrum characteristics. These templates serve as auxiliary references during the diagnostic process. When the confidence rule base inference results are uncertain, correlation matching is performed between the real-time time-frequency feature matrix and the templates to further assist in determining the fault type.

[0064] During system integration, the confidence rule base reasoner interacts with the feature extraction module, the condition assessment module, and the optimization and decision-making module. The time-frequency feature matrix generated by the feature extraction module provides input data for the fault mode knowledge base. The bearing health probability value output by the condition assessment module serves as one of the parameters for the activation module. The optimal feature combination data selected by the optimization and decision-making module is used to improve the validity of the input parameters and the accuracy of the diagnosis. Through the collaborative work of these modules, a complete bearing operating condition detection and diagnosis process is formed.

[0065] Example 4: During data preprocessing for a multi-channel sensor module, different types of sensor data must be processed using corresponding algorithms to extract effective features and suppress noise, providing a reliable data foundation for subsequent status assessment and diagnosis. The following describes in detail the preprocessing implementation for each sensor data, using specific examples:

[0066] Signals collected by vibration accelerometers typically contain a wealth of fault signatures, but they are also susceptible to interference from mechanical vibration background noise. Therefore, an envelope demodulation algorithm is required to extract the characteristic frequency components of the fault. For example, in the case of a bearing pitting fault, suppose a vibration acceleration time domain signal is collected at a certain moment. The original waveform is mixed with high-frequency noise and low-frequency mechanical vibration interference. The preprocessing steps are as follows: A high-pass filter is used to remove low-frequency interference. The high-pass filter's cutoff frequency is set to 500Hz to filter out noise generated by low-frequency mechanical vibrations, such as gear meshing, while retaining high-frequency signals that may contain bearing fault signatures (such as impact vibration signals from bearing components). Although the filtered signal still contains high-frequency noise, the frequency components associated with the bearing fault have been initially isolated.

[0067] The filtered signal is demodulated using the Hilbert envelope. The Hilbert transform converts a real signal into an analytical signal and extracts its envelope, thereby highlighting the periodic impact components within the signal. For example, pitting faults can cause periodic impact vibrations in a bearing during rotation, and the envelope signal will exhibit periodic fluctuations related to the fault's characteristic frequency. By calculating the power spectral density of the envelope signal, a clear peak frequency can be observed in the spectrum, corresponding to the characteristic frequency of bearing pitting faults (e.g., 1.5 times the rotational frequency). This process allows fault characteristics hidden within complex vibration signals to be clearly extracted, facilitating subsequent feature analysis and fault diagnosis.

[0068] The current sensor monitors the motor input circuit's current signal in real time. Changes in its waveform reflect changes in bearing load and the motor's operating status. For example, in the case of increased load due to bearing wear, the collected current signal may contain both a fundamental component and higher-order harmonics. An increase in higher-order harmonics often indicates a mechanical failure. The preprocessing steps are as follows: A fast Fourier transform (FFT) is used to convert the current time-domain signal into a frequency domain signal, generating a spectrum. The FFT point count is set to 4096 to ensure the frequency resolution meets diagnostic requirements (frequency resolution = sampling frequency / number of FFT points). Assuming a sampling frequency of 10 kHz, the frequency resolution is approximately 2.44 Hz, enabling clear discrimination of the various harmonic components in the current signal (e.g., a 50 Hz fundamental, a 150 Hz third harmonic, a 250 Hz fifth harmonic, and so on).

[0069] Background noise is suppressed through adaptive threshold filtering. The threshold is calculated based on the mean power of the noise segment in the spectrum. The specific method is as follows: select a frequency band in the spectrum with no obvious signal components (such as a high-frequency noise band), calculate the mean power spectral density within this frequency band, and use 1.5 times this mean as the cutoff threshold. Frequency components below the threshold are considered background noise and set to zero; frequency components above the threshold are retained as effective features reflecting the bearing condition. For example, under normal operating conditions, the harmonic components of the current signal are mainly fundamental waves, and the power of higher-order harmonics is relatively low. When bearing wear causes uneven load, significant higher-order harmonics (such as third and fifth harmonics) appear in the current signal, and their power spectral density is higher than the threshold. Filtering can highlight these features, providing a basis for determining bearing load anomalies.

[0070] A temperature sensor (such as a PT100 thermal resistor) is embedded in the bearing lubrication cavity to monitor the bearing's operating temperature. Due to sensor measurement errors and environmental fluctuations, the raw temperature data may fluctuate randomly. An exponentially weighted moving average (EWMA) algorithm is used to eliminate measurement fluctuations. For example, consider the temperature monitoring of a bearing during continuous operation. Assuming that temperature data is collected every minute, a series of temperature values ​​are obtained. .

[0071] The EWMA algorithm generates a smoothed temperature value by taking a weighted average of the current measurement value and the filter result of the previous moment. The specific calculation method is: the current moment's filter value is equal to the smoothing coefficient multiplied by the current measurement value, plus (1-smoothing coefficient) multiplied by the filter value of the previous moment. The smoothing coefficient is usually set to 0.2, which gives the recent measurement value a higher weight, can respond to the temperature change trend more quickly, and suppress the impact of short-term fluctuations. For example, if the first The measured temperature in minutes is , No. The filtered value for the minute is , then The filtered value for the minute is In this way, random noise in the temperature data is effectively smoothed, and temperature change trends (such as slow increases or decreases) are clearly presented, making it easier to analyze whether the thermal status of the bearing is normal.

[0072] The speed encoder is rigidly connected to the drive shaft via an incremental code disc and outputs a pulse signal that reflects the bearing speed. Due to mechanical transmission backlash and measurement errors, the directly acquired pulse signal may exhibit jitter. Therefore, a Kalman filter algorithm is used to fuse the angular velocity and angular acceleration data to output the bearing's instantaneous speed and speed fluctuation rate. For example, consider the speed variation of a gearbox bearing during a gear shift. Assuming a 10ms sampling period for the speed encoder, the angular velocity (in rad / s) and angular acceleration (in rad / s²) can be calculated within each sampling period.

[0073] The Kalman filter algorithm uses angular velocity and angular acceleration as state variables, constructing state equations and observation equations. The state equation describes how the state variable changes over time; for example, the derivative of angular velocity is angular acceleration. The observation equation describes the relationship between the observed value (such as the angular velocity calculated by pulse counting) and the state variable. By recursively calculating the state estimate and covariance matrix, the Kalman filter effectively integrates measurement data from multiple moments, suppresses noise interference, and outputs smooth instantaneous speed and speed fluctuation. For example, when a bearing switches from high speed (2000 rpm) to low speed (1000 rpm), the original pulse signal may experience brief jitter due to the gear shift shock. After Kalman filtering, the instantaneous speed curve transitions smoothly, and the speed fluctuation (such as the standard deviation of the speed change per second) accurately reflects the severity of the speed change, providing data support for determining whether the bearing operating condition transition is smooth.

[0074] In practical applications, the data from each sensor in a multi-channel sensing module must be processed collaboratively to ensure data consistency and reliability. For example, vibration acceleration data is strongly correlated with speed encoder data, and changes in bearing speed directly affect the frequency components of the vibration signal (such as the rotational frequency and its multiples). Therefore, during preprocessing, signal anomalies can be detected by calculating the correlation coefficient (such as the Spearman correlation coefficient) between the data from each channel. If the correlation coefficient between the data from a certain channel and the associated channel falls below a set threshold (such as 0.5), the channel data is considered to have a possible anomaly (such as a sensor failure or loose wiring), and data reconstruction is required. Data reconstruction can leverage the correlation between historical data and adjacent channel data to generate alternative data through linear interpolation or regression models to ensure the continuity of subsequent analysis.

[0075] For example, consider a vibration accelerometer channel anomaly. Suppose, at a certain moment, the vibration signal's correlation coefficient suddenly drops to 0.3, while speed, current, and temperature data are all normal. This indicates a possible vibration sensor failure. The system automatically retrieves the previous 10 minutes of vibration and speed data, builds a linear regression model (e.g., the linear relationship between vibration RMS value and speed), and uses the current speed value to predict an estimated vibration signal value, replacing the abnormal data. This approach allows the system to maintain operation based on data from other sensors, even if a single sensor fails, avoiding diagnostic interruptions due to missing data.

[0076] Data preprocessing for the multi-channel sensor module effectively improves data quality and feature recognition through algorithm design tailored to the characteristics of different sensors. Envelope demodulation of vibration data, spectrum analysis of current data, smoothing of temperature data, and Kalman filtering of speed data extract key information about bearing operation from different dimensions. Correlation verification and reconstruction mechanisms for multi-sensor data further enhance system robustness. These preprocessing steps lay a solid foundation for subsequent feature extraction, state assessment, and fault diagnosis, ensuring accurate capture of bearing operating state changes under active-passive switching conditions, enabling early fault warning and precise maintenance.

[0077] Example 5: In the implementation process of generating a multi-level diagnostic strategy at the analysis layer of the hierarchical diagnostic control model, it is necessary to combine the bearing health status probability value, state transition rules and trend prediction requirements to build a diagnostic system covering different risk levels. The following describes the implementation method in detail with reference to specific working conditions:

[0078] 1. Health Level Classification and Response Mechanism

[0079] Bearing health status is categorized into three levels: "Normal," "Warning," and "Fault," each corresponding to a different probability threshold and response strategy. For example, in the Normal level, a health probability value ≥ 0.7 indicates that the bearing is in good operating condition. This triggers periodic monitoring, set to a 30-minute interval. The system only records trend data for key parameters such as vibration energy entropy and current harmonic distortion, without performing real-time in-depth analysis. For example, when a bearing is operating at high speed and light load, the health probability value is 0.85. The system collects and stores this data every 30 minutes for long-term trend analysis.

[0080] Warning Level: A healthy probability value between 0.4 and 0.7 indicates the possibility of early bearing anomalies. At this point, feature tracking monitoring is triggered, the monitoring cycle is shortened to 5 minutes, and a multi-feature joint analysis process is initiated. For example, when the bearing healthy probability value drops to 0.6, the system automatically increases the frequency of analysis of features such as the vibration signal's envelope spectrum and temperature gradient, comparing them with historical data to determine whether the abnormal trend persists.

[0081] Fault Level: A healthy probability value of 0.4 or less indicates a significant bearing failure or impending failure. This triggers the shutdown protection protocol, sending a shutdown command to the transmission controller via the CAN bus to prevent further failure. For example, if the healthy probability value drops sharply to 0.2, the system immediately shuts off motor power to prevent bearing locking and severe transmission damage.

[0082] 2. State Transition Matrix and Upgrade Diagnosis Process

[0083] A state transition matrix is ​​constructed to describe the change of health level over time. The matrix elements are set based on historical data statistics or expert experience. For example, the state transition matrix can be expressed as: Among them, the first row indicates that the probabilities of transitioning from the "normal" level to "normal", "warning", and "fault" are 90%, 9%, and 1% respectively; the second row indicates that the probabilities of transitioning from the "warning" level to each state are 10%, 80%, and 10% respectively; the third row indicates that the probabilities of transitioning from the "fault" level to each state are 0%, 5%, and 95% respectively.

[0084] When the health level decreases (e.g., "normal" → "warning" → "fault") for three consecutive sampling periods (assuming a sampling period of 1 second), the upgraded diagnostic process is triggered. For example, if the health level of a bearing is "normal" at time t1, drops to "warning" at time t2, and further drops to "fault" at time t3, the system will determine that the condition is continuously deteriorating and automatically activate the backup sensor channel (such as a redundant vibration accelerometer) to perform data verification, comparing the vibration data from the main and backup channels to eliminate misjudgments caused by a single sensor failure. If the verification confirms that the level decline trend is consistent, the fault level response measures are implemented; if the data is inconsistent, a sensor anomaly warning is issued, and hardware maintenance is prioritized.

[0085] 3. Trend Prediction Mechanism and Active Testing Mode

[0086] A trend prediction mechanism is introduced. When the rate of change of the bearing health probability value exceeds a preset gradient threshold (e.g., an absolute value of 0.05 / second), an active test mode is initiated. For example, if a bearing's health probability value suddenly decreases at a rate of 0.06 / second during normal operation, the system may identify a rapidly developing fault (e.g., a bearing crack propagation) and immediately trigger the active test mode. A step load excitation (e.g., from 50% to 80% of the rated load) is applied through the transmission controller. The system then collects dynamic bearing response data (e.g., vibration and current signals) for in-depth time-frequency analysis and fault signature matching.

[0087] In active testing mode, the system can capture fault characteristics that are not obvious under static conditions. For example, an early-stage crack in a bearing may only cause slight vibration under static light load. However, under a step load, the crack's deformation produces significant impact vibration, with a significant increase in the kurtosis value and characteristic frequency components (such as twice the rotational frequency) of the vibration signal. By comparing the data changes before and after the excitation, the fault type and severity can be more accurately determined, providing a basis for formulating maintenance strategies.

[0088] 4. Constraints of the Multi-parameter State Assessment Model

[0089] Multi-parameter state assessment models must meet the following constraints to ensure engineering practicality: Feature dimension constraint: This limits the maximum dimension of the feature vector to the upper threshold of computing resource allocation (e.g., 12 dimensions). For example, the raw features extracted from vibration, current, temperature, speed, and other data may exceed 20 dimensions (e.g., vibration root mean square, kurtosis, margin index, and current harmonic energy). Principal component analysis (PCA) or quantum genetic optimization algorithms must be used to select the most representative 12-dimensional features (e.g., vibration energy entropy, current harmonic distortion rate, temperature gradient, etc.) to avoid excessive computational complexity due to too many features, which could affect system real-time performance.

[0090] Operating condition coverage constraints: The adaptability of the feature combination is tested by simulating four typical operating conditions: high speed, low speed, heavy load, and light load. For example, under high speed and light load conditions, the bearing primarily exhibits high-frequency vibration, so the feature combination must include parameters that reflect high-frequency impacts (such as the high-frequency energy of the vibration signal). Under low speed and heavy load conditions, bearing temperature changes and current fluctuations are more significant, so the feature combination must focus on the stability of the temperature gradient and the fundamental current component. Through multi-condition testing, we ensure that the feature combination can effectively distinguish between healthy and faulty states under different operating conditions.

[0091] Real-time constraints: The maximum latency of a single diagnostic process is limited to no more than 20% of the control period (e.g., if the control period is 100ms, the diagnostic latency is ≤20ms). To meet this requirement, algorithm execution efficiency must be optimized. For example, a parallel computing architecture can be used for the variational mode decomposition algorithm and hardware acceleration can be implemented for the probabilistic neural network model. This ensures that the entire process, from data acquisition to diagnostic output, is completed within the specified timeframe, avoiding delays that can lead to delayed fault responses.

[0092] 5. Signal Verification Module and Data Reconstruction

[0093] A signal verification module was built to detect signal anomalies by calculating the correlation coefficient of multi-sensor data. For example, vibration acceleration data and rotational speed data are typically strongly correlated (e.g., the vibration characteristic frequency is a multiple of the rotational speed). If the Spearman correlation coefficient between the vibration data and the rotational speed data for a particular channel falls below 0.5 (a preset threshold), the vibration data is considered abnormal. At this point, data reconstruction is initiated, using the linear relationship between historical vibration data and rotational speed data (e.g., the RMS value of vibration increases with increasing rotational speed) to generate surrogate data using linear interpolation.

[0094] For example, if the speed encoder fails, the system might detect a sudden drop in the correlation between the vibration and speed data, while the vibration, current, and temperature data are normal. This can lead to a sudden drop in the speed data, leading to a speed anomaly. Based on the speed trend over the previous five seconds (e.g., a steady decrease) and the current motor current (which is negatively correlated with speed), the system estimates the current speed value and reconstructs the data, maintaining continuity in the diagnostic process. The system also issues a speed sensor failure warning, prompting maintenance personnel to conduct prompt repairs.

[0095] 6. Diagnostic Process Examples under Typical Working Conditions

[0096] The practical application of the hierarchical diagnostic control model is illustrated using the example of an AMT transmission bearing switching from active (motor-driven) to passive (coasting) operation. During the active phase, the bearing rotates at high speed driven by the motor, and the health probability is 0.75 (the lower limit of the warning level). The system triggers feature tracking monitoring, analyzing the time-frequency characteristics of the vibration signal every five minutes. Energy concentration at 0.8 times the rotational frequency (corresponding to a wear fault) is detected in the vibration signal, but the health probability has not yet exceeded the fault threshold. The system maintains the warning level and records the trend of feature changes.

[0097] During the operating mode transition phase, after the transmission shifts into the passive mode, the bearing speed decreases, and the health probability value suddenly drops at a rate of 0.07 / second to 0.38 (fault level). Because the rate of change exceeds the threshold, the system initiates active testing mode, applying a short-term load excitation. The collected vibration signal shows a significant increase in the double-speed frequency component (corresponding to a crack fault signature). Combined with the high transition probability (10%) from "warning to fault" in the state transition matrix, the bearing is determined to be at risk of crack propagation.

[0098] Fault response phase: The system immediately triggered the shutdown protection protocol and, through the confidence rule base reasoner, output a Level III warning signal and an "immediate shutdown for repair" maintenance recommendation. After inspection, a deep crack was found in the bearing inner race, confirming the accuracy of the diagnosis.

[0099] The hierarchical diagnostic control model achieves dynamic tracking and precise response to bearing conditions through multi-level diagnostic strategies, state transition analysis, trend prediction, and constraint management. Differentiated health level monitoring, continuity assessment of state transitions, fault signature mining through active testing, and feature optimization under multiple constraints ensure the system's reliability and real-time performance under active-passive switching conditions. Signal verification and data reconstruction mechanisms further enhance the system's fault tolerance, enabling stable operation in complex industrial environments and providing effective technical support for the full lifecycle health management of AMT transmission bearings.

[0100] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0101] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting the operating status of an AMT bearing under active and passive switching conditions, characterized in that: include: Collecting real-time operating data of the gearbox bearings through a multi-channel sensing module, which includes a vibration acceleration sensor, a current sensor, a temperature sensor, and a speed encoder; The real-time operating data is extracted for intrinsic mode components based on a variational mode decomposition algorithm, and a time-frequency feature matrix is ​​generated through a Hilbert transform. The time-frequency feature matrix is ​​input into a pre-trained probabilistic neural network model. The probabilistic neural network model uses a sliding time window mechanism to adjust the hidden layer node activation threshold based on the real-time operating condition data and outputs a bearing health status probability value. A multi-parameter state assessment model is constructed based on the bearing health status probability value. The multi-parameter state assessment model optimizes the highest fault sensitivity and lowest feature redundancy and performs a global search for feature subsets using a quantum genetic optimization algorithm, wherein the quantum genetic optimization algorithm introduces adaptive rotation angle and dynamic crossover probability. Optimal feature combination data is output based on the multi-parameter state assessment model. A hierarchical diagnostic control model is established based on the optimal feature combination data. The hierarchical diagnostic control model includes an acquisition layer, an analysis layer, and an execution layer. The acquisition layer updates the operating mode based on the time-frequency feature matrix, the analysis layer generates a multi-level diagnostic strategy based on the bearing health status probability value, and the execution layer maps diagnostic instructions into warning levels and maintenance recommendation signals using a confidence rule base reasoner.

2. The method for detecting the operating status of an AMT bearing under active-passive switching conditions according to claim 1 is characterized in that: The probabilistic neural network model adopts a sliding time window mechanism, including: constructing a probabilistic network structure, the probabilistic network structure includes a vibration feature node, a current feature node and a health status node, the vibration feature node and the current feature node are parent nodes, and the health status node is a child node; updating the probability density function of the vibration feature node based on real-time collected data, performing frequency band energy analysis on the current feature node through a wavelet packet decomposition algorithm, and obtaining the posterior probability of the characteristic frequency band energy distribution; introducing a dynamic threshold adjustment mechanism, dynamically adjusting the inference step size of the probabilistic network according to changes in the bearing load, and the time window width is negatively correlated with the bearing speed; using an expectation maximization algorithm to perform online parameter update on the probabilistic neural network, and optimizing the hidden layer node weight parameters by maximizing the likelihood function.

3. The method for detecting the operating status of an AMT bearing under active-passive switching conditions according to claim 1 is characterized in that: The quantum genetic optimization algorithm introduces adaptive rotation angle and dynamic crossover probability, including: constructing a genetic optimization objective function, the objective function includes a feature sensitivity term and a feature independence term, the feature sensitivity term is calculated by the inter-class distance of the fault sample, and the feature independence term is calculated by the correlation coefficient between the feature vectors; designing an adaptive rotation angle update rule, the rotation angle changes in an exponential decay curve with the increase of evolutionary generations, maintaining a large search range in the initial stage and converging to the optimal solution neighborhood in the later stage; designing a dynamic crossover probability adjustment mechanism, the crossover probability is dynamically adjusted according to the population fitness variance, and the crossover operation frequency is increased when the population variance is higher than the threshold, and vice versa.

4. The method for detecting the operating status of an AMT bearing under active and passive switching conditions according to claim 1 is characterized in that: The confidence rule base reasoner maps diagnostic instructions into warning levels and maintenance recommendation signals, including: constructing an input activation module, wherein the input includes the probability value of the bearing health status, the temperature change gradient, and the bearing operating time, and using a Gaussian membership function to assign confidence to the input variables; constructing a rule base, wherein the rule base contains confidence rules; designing an evidence fusion module, which uses DS evidence theory to fuse multi-source rule outputs into a comprehensive confidence distribution, wherein the output includes a warning level code and a maintenance recommendation code; and introducing a rule weight correction mechanism to dynamically update the rule confidence coefficient based on historical diagnostic accuracy and maintenance records.

5. The method for detecting the operating status of an AMT bearing under active-passive switching conditions according to claim 4 is characterized in that: The rule construction of the confidence rule base reasoner includes: establishing a fault mode knowledge base, which contains vibration spectrum feature templates of three typical faults of bearing pitting, cracking and wear.

6. The method for detecting the operating status of an AMT bearing under active and passive switching conditions according to claim 1 is characterized in that: The data preprocessing of the multi-channel sensing module includes: using an envelope demodulation algorithm to extract fault characteristic frequency components from vibration acceleration data; using fast Fourier transform to generate a spectrum diagram for current sensor data, and suppressing background noise through adaptive threshold filtering; using an exponentially weighted moving average algorithm to eliminate measurement fluctuations for temperature sensor data; and using a Kalman filter algorithm to fuse angular velocity and angular acceleration data for speed encoder data, and outputting the instantaneous speed and speed fluctuation rate of the bearing.

7. The method for detecting the operating status of an AMT bearing under active and passive switching conditions according to claim 1 is characterized in that: The analysis layer of the hierarchical diagnostic control model generates a multi-level diagnostic strategy, including: dividing the health level into three levels: normal, warning, and fault. The normal level triggers periodic monitoring, the warning level triggers feature tracking monitoring, and the fault level triggers the shutdown protection protocol; constructing a state transition matrix, which performs state jumps according to continuous health level changes. If the health level decreases for three consecutive sampling periods, an upgraded diagnostic process is triggered; introducing a trend prediction mechanism, and starting the active test mode when the change rate of the bearing health status probability value exceeds a preset gradient threshold.

8. The method for detecting the operating status of an AMT bearing under active and passive switching conditions according to claim 1 is characterized in that: The constraints of the multi-parameter state assessment model include: feature dimension constraint, which limits the maximum dimension of the feature vector to no more than the upper threshold of computing resource allocation; operating condition coverage constraint, which verifies the operating condition adaptability of the feature combination through typical operating condition simulation tests; and real-time constraint, which limits the maximum delay of a single diagnostic process to no more than 20% of the control period.

9. The method for detecting the operating status of an AMT bearing under active and passive switching conditions according to claim 1 is characterized in that: The method further includes: constructing a signal verification module to detect signal anomalies through multi-sensor data correlation analysis, and initiating data reconstruction if the correlation coefficient between a channel data and an associated channel is lower than a set threshold.

10. An AMT bearing operating status detection system under active and passive switching conditions, characterized in that: include: Multi-channel sensing module: used to collect real-time operating data of gearbox bearings, including vibration acceleration sensor, current sensor, temperature sensor and speed encoder; Feature extraction module: Based on the variational mode decomposition algorithm, the intrinsic mode components of the real-time operating data collected by the multi-channel sensing module are extracted, and the time-frequency feature matrix is ​​generated through Hilbert transform; state assessment module: The time-frequency feature matrix is ​​input into the pre-trained probabilistic neural network model. The probabilistic neural network model adopts a sliding time window mechanism, adjusts the hidden layer node activation threshold based on the real-time operating condition data, and outputs the bearing health status probability value; optimization decision module: Constructs a multi-parameter state assessment model based on the bearing health status probability value, takes the highest fault sensitivity and the lowest feature redundancy as the optimization goal, and uses a quantum genetic optimization algorithm that introduces adaptive rotation angle and dynamic crossover probability to perform a global search on the feature subset to output the optimal feature combination data; diagnosis execution module: Based on the optimal feature combination data, it is established, including the acquisition layer, analysis layer and execution layer. The acquisition layer updates the operating mode based on the time-frequency feature matrix, the analysis layer generates a multi-level diagnosis strategy based on the bearing health status probability value, and the execution layer maps the diagnostic instructions to warning levels and maintenance recommendation signals through the confidence rule base reasoner.

Images