Seat fault diagnosis method based on magneto-rheology

Abstract

The invention relates to the technical field of magneto-rheological seat diagnosis, and discloses a fault diagnosis method based on a magneto-rheological seat. The method comprises the following steps: acquiring current state data of the magnetorheological seat, and establishing a fault feature database mapped with the current state data; configuring diagnosis targets according to the fault feature database, wherein the diagnosis targets comprise vibration abnormity diagnosis, pressure distribution diagnosis and response time diagnosis; then determining a target classification identifier in the diagnosis target, wherein the target classification identifier comprises a main diagnosis target, an auxiliary diagnosis target and a reference target; after key diagnosis parameters are selected through the main diagnosis target and the auxiliary diagnosis target, diagnosis target influence analysis of the key diagnosis parameters is executed; and finally, after parameter optimization constraints are established according to the diagnosis target influence analysis result, parameter directional optimization is executed, and magneto-rheological seat fault diagnosis is completed by using a parameter directional optimization result. According to the method, the diagnosis efficiency and accuracy can be improved, and stable operation of the magneto-rheological seat is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of magnetorheological seat diagnostic technology, specifically to a fault diagnosis method for magnetorheological seats. Background Technology

[0002] Magnetorheological (MR) seats, as a type of seating system integrating advanced magnetorheological technology, have been widely used in the automotive, aerospace, and high-end medical equipment industries due to their ability to dynamically adjust vibration damping, pressure distribution, and response speed according to usage scenarios. With the increasing complexity of usage scenarios and the growing demands for seat comfort and safety, fault diagnosis of MR seats has become a crucial aspect of ensuring their stable operation. Fault diagnosis for magnetorheological (MR) seats often employs traditional single-parameter testing methods, which determine the presence of a fault by testing a single performance parameter of the seat. This approach has significant limitations: firstly, the various performance parameters of a MR seat are interconnected; for example, abnormal vibration may simultaneously lead to uneven pressure distribution, and single-parameter testing cannot capture this correlation, easily resulting in misdiagnosis or missed diagnosis. Secondly, traditional diagnostic methods do not clearly classify and prioritize diagnostic targets, making it difficult to quickly pinpoint the core cause of the fault when faced with multiple parameter anomalies, leading to low diagnostic efficiency. Current diagnostic technologies lack a systematic fault characteristic database, requiring data to be re-collected and analyzed for each diagnosis, hindering the reuse of historical data and the accumulation of diagnostic experience. Furthermore, the diagnostic process fails to consider parameter optimization constraints, relying solely on fixed thresholds to determine faults. This fails to adapt to the dynamic changes in the performance parameters of magnetorheological (MR) seats under different usage environments, significantly impacting the accuracy and applicability of diagnostic results. These issues not only increase the maintenance costs of MR seats but also potentially affect safety and comfort due to untimely and inaccurate fault identification, thus restricting the further promotion and application of MR seat technology. Summary of the Invention

[0003] The purpose of this invention is to provide a fault diagnosis method for a magnetorheological seat to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a fault diagnosis method for a magnetorheological seat, the method comprising: Acquire the current state data of the magnetorheological seat and establish a fault feature database that maps to the current state data; Diagnostic targets are configured according to the fault feature database, including vibration anomaly diagnosis, pressure distribution diagnosis, and response time diagnosis. Determine the target classification identifiers in the diagnostic objectives, wherein the target classification identifiers include primary diagnostic objectives, auxiliary diagnostic objectives, and reference objectives; After selecting key diagnostic parameters using the primary and secondary diagnostic objectives, an impact analysis of the diagnostic objectives on the key diagnostic parameters is performed. After establishing parameter optimization constraints based on the results of the diagnostic target impact analysis, parameter-oriented optimization is performed, and the magnetorheological seat fault diagnosis is completed using the results of parameter-oriented optimization.

[0005] Preferably, acquiring the current state data of the magnetorheological seat includes: When the preset diagnostic cycle is met, a retrospective profile of the acquisition time of the target sensor data of the magnetorheological seat is obtained, wherein the retrospective profile includes a first normal state value and a first control parameter; A real-time profile of the magnetorheological seat is obtained, wherein the real-time profile includes a second current state value and a second control parameter.

[0006] Preferably, the retrospective profile of the acquisition time of the target sensor data of the magnetorheological seat includes: Multiple preset control parameters are used to obtain the target sensor data; With each preset control parameter as a constraint, the alienation state value of the target sensor data is statistically analyzed based on the historical sample library; Extract 0.5 times the minimum value of all alienated state values ​​and set it as the preset diagnostic cycle.

[0007] Preferably, the impact analysis of the diagnostic target for performing key diagnostic parameters includes: The time delay between the signals from the two monitoring points is determined based on the correlation between the sensor signals at the two monitoring points. The signal fluctuation factor is determined based on the amplitude distribution of the sensor signal at the monitoring point within a short time window and the amplitude distribution of the sensor signal at the next monitoring point after a time delay within the short time window. Based on the amplitude difference, phase difference, and signal fluctuation factor of the sensor signals contained in all short-time windows of adjacent monitoring points in the frequency domain, the interference probability of the sensor signals at each frequency of the monitoring points is determined. The interference frequency is obtained based on the interference probability of the sensor signals at the monitoring point at all frequencies; The nonlinear error probability is obtained based on the distribution of interference frequencies at all adjacent monitoring points.

[0008] Preferably, the step of establishing parameter optimization constraints based on the impact analysis results of the diagnostic target includes: Retrieve the set of target sensor data deviation magnitude values ​​that satisfy the profile deviation vector. If the variance of the set of target sensor data deviation magnitude values ​​is greater than or equal to the variance threshold, calculate the failure probability. Parameter optimization constraints are established based on the aforementioned failure probabilities.

[0009] Preferably, the execution parameter orientation optimization includes: After establishing the control range of the parameters, an initial solution set is created based on the current parameter set; After performing fitness evaluation of solutions within the initial solution set, the optimization direction and optimization step size are established through parameter optimization constraints and fitness evaluation results; The initial solution set is iteratively updated using the optimization direction and the optimization step size. The parameter optimization is completed based on the iterative update results.

[0010] Preferably, the iterative update of the initial solution set using the optimization direction and the optimization step size includes: For each solution, an iterative trajectory is established, and the iterative trajectory is identified by the fitness value of the solution in each iteration. Configure an iterative evaluation interval, identify the update status of the iterative trajectory within the iterative evaluation interval, and generate an evaluation classification, which includes a good solution evaluation classification, an exploration evaluation classification, and a poor solution evaluation classification. Search self-optimization management is performed through iterative updates based on the evaluation classification.

[0011] Preferably, the search self-optimization management that iteratively updates based on the evaluation classification includes: In the optimal solution evaluation classification configuration, a local proxy model is used to predict the improvement trend and generate a first reference optimization direction. In the inferior solution evaluation classification configuration, a penalty optimization identification layer is configured, and the penalty optimization identification layer is used to identify the direction of error improvement and establish window improvement taboos; The solution fine-tuning iteration update within the optimal solution evaluation category is performed using the first reference optimization direction and window improvement taboo, and the hybrid exploration iteration update within the exploration evaluation category is performed using the first reference optimization direction and window improvement taboo. A random factor is configured to perform the iteration update of the solution within the inferior solution evaluation category.

[0012] Preferably, the execution parameter-oriented optimization further includes: Establish an energy consumption evaluation function for the parameters; Establish a balance adaptation function based on the energy consumption evaluation function and the quality evaluation function; Based on the aforementioned equilibrium fitness function, the optimization scheme is screened for parameter-oriented optimization, and the optimization scheme screening result is output as the parameter-oriented optimization result.

[0013] Preferably, the process of using parameter-oriented optimization results to complete the fault diagnosis of the magnetorheological seat includes determining the fault type and fault severity based on the parameter-oriented optimization results.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This fault diagnosis method for magnetorheological seats achieves systematic integration and management of operating status data by establishing a fault feature database mapped to current state data. The database eliminates the need for extensive re-collection of basic data for each diagnosis; historical data can be directly compared and analyzed with current data, reducing redundant data processing steps. Simultaneously, it continuously accumulates characteristic data corresponding to different fault types, providing richer reference data for subsequent diagnoses and effectively improving the coherence of the diagnostic process and data utilization efficiency. In the diagnostic target configuration phase, the diagnostic targets are clearly divided into vibration anomaly diagnosis, pressure distribution diagnosis, and response time diagnosis, covering the core performance dimensions of the magnetorheological seat. This ensures that the diagnostic process can comprehensively capture potential fault points in the seat and avoids fault omissions due to missing diagnostic dimensions. By defining target classification labels to distinguish between primary diagnostic targets, auxiliary diagnostic targets, and reference targets, the diagnostic focus can be quickly identified when multiple parameters are abnormal simultaneously. Priority can be given to in-depth analysis of parameters related to the primary diagnostic target, avoiding wasting time on non-core parameters, significantly improving the efficiency of fault location, and helping diagnostic personnel focus on key issues more quickly. By using the primary and secondary diagnostic objectives to select key diagnostic parameters and then performing the diagnostic objective impact analysis, the correlation between different key parameters and each diagnostic objective can be clearly identified. This allows for understanding the specific impact of an abnormal parameter on different performance dimensions such as vibration, pressure distribution, or response time. This breaks the limitation of isolated parameters in traditional single-parameter diagnosis, making the diagnostic process more logical and systematic, and reducing misjudgments caused by unclear parameter correlations. Based on the impact analysis results of diagnostic objectives, parameter optimization constraints are established and parameter-oriented optimization is performed, making the parameter optimization process more targeted and reasonable. This process fully considers the interaction between different diagnostic objectives and the parameter adaptation requirements in actual use scenarios, avoiding the blindness of parameter optimization under unconstrained conditions. It can find a parameter range that better matches the actual operating state of the seat. The diagnostic results based on this parameter range are more consistent with the actual fault conditions of the seat, effectively improving the accuracy and reliability of the diagnostic results and reducing diagnostic bias caused by unreasonable parameter optimization. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating the working principle of the magnetorheological seat fault diagnosis method described in this invention. Figure 2 A flowchart for obtaining a profile by retrospectively analyzing data collected from the target sensor. Figure 3 A flowchart for the impact analysis of key diagnostic parameters on diagnostic targets; Figure 4 The flowchart for parameter-oriented optimization. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 This invention provides a fault diagnosis method for a magnetorheological seat, the method comprising: Acquire the current state data of the magnetorheological seat and establish a fault feature database mapped to the current state data; configure diagnostic targets based on the fault feature database, including vibration anomaly diagnosis, pressure distribution diagnosis, and response time diagnosis; determine the target classification identifiers in the diagnostic targets, including primary diagnostic targets, auxiliary diagnostic targets, and reference targets; select key diagnostic parameters using the primary and auxiliary diagnostic targets, and then perform diagnostic target impact analysis on the key diagnostic parameters; establish parameter optimization constraints based on the results of the diagnostic target impact analysis, and then perform parameter-oriented optimization; use the results of parameter-oriented optimization to complete the fault diagnosis of the magnetorheological seat.

[0018] Example 1: See Figure 2 In the implementation of the magnetorheological seat fault diagnosis method, acquiring current state data and establishing a retrospective profile are fundamental and crucial steps. This process begins with continuous monitoring of the seat system's built-in sensors, which typically include accelerometers, pressure sensors, displacement sensors, and control current monitoring modules. These sensors respectively collect the seat's vibration acceleration, pressure distribution on the seat surface and backrest, actuator displacement, and the input current signal of the magnetorheological damper. The data acquisition system operates at a fixed sampling frequency, and the raw data is filtered and preprocessed before being stored in a circular buffer.

[0019] The determination of the preset diagnostic cycle is a dynamic calculation process. The system accesses a historical sample database, which stores massive amounts of data on the seat's long-term operation under known normal conditions. For each preset control parameter, such as vibration amplitude threshold, pressure equalization, or current response range, the system statistically analyzes the historical anomaly values ​​of the corresponding target sensor data. Anomaly values ​​refer to the degree to which sensor readings deviate from their normal fluctuation range, and their calculation relies on the statistical distribution characteristics of historical data. The system iterates through all important control parameters, calculates their corresponding anomaly value sequences, and finds the minimum value. Fifty percent of this minimum value is set as the current diagnostic cycle. This means that the system's diagnostic trigger frequency adaptively adjusts based on the potential risks of data anomalies. When historical data shows the system is in a relatively stable period, the diagnostic cycle is extended accordingly to save computational resources; conversely, when data shows minor anomalies, the diagnostic cycle is automatically shortened to increase monitoring frequency.

[0020] When the dynamically calculated diagnostic cycle is met, the system triggers the generation of a retrospective profile. This profile is not a simple snapshot of historical data, but rather a feature set constructed based on the historical context information of the target sensor data at a specific acquisition moment. The system locks onto that acquisition moment and extracts all relevant sensor data from the historical sample library within a time window before and after that moment. The first normal state value is a statistical feature extracted from the historical window data, such as the root mean square value of the vibration signal, the average and variance of the pressure distribution, and the average response time. These values ​​represent the normal baseline that the system should be in at that specific moment. The first control parameter is the set of control commands actively applied by the system within that historical time period, such as the current command value provided to the damper and the target position of the actuator. The essence of the retrospective profile is to establish a historical reference baseline for real-time diagnosis.

[0021] Simultaneously, real-time profiling is generated. This real-time profiling is based on the real-time sensor data stream within the latest time window. Using the same feature extraction method as the retrospective profiling, a second current state value and a second control parameter are obtained. The second current state value reflects the system's current real-time operating characteristics, while the second control parameter records the control commands currently being executed by the system. Comparing the real-time profiling with the retrospective profiling forms the basis for subsequent fault feature database establishment and diagnosis. For example, vibration anomaly diagnosis might focus on the deviation between the feature values ​​of the real-time vibration signal and the baseline values ​​in the retrospective profiling; pressure distribution diagnosis would compare the uniformity index of the real-time pressure distribution with the historical baseline; and response time diagnosis would compare the current system's response delay to control commands with the historical normal response time. Through this profiling comparison based on historical and real-time data, the system can more accurately capture potential fault symptoms, rather than simply relying on fixed threshold judgments.

[0022] The entire data acquisition and profiling process is completed on an embedded data processing unit. This unit has sufficient data caching and computing power to process multi-channel sensor signals in real time and perform historical data queries and feature calculations. The historical sample library is typically stored in non-volatile memory and is periodically updated to include new normal operating condition data to ensure the representativeness of its baseline.

[0023] Example 2: See Figure 3 In the fault diagnosis of magnetorheological (MRT) seats, the impact analysis of diagnostic targets for key diagnostic parameters is a crucial step in deeply exploring the internal signal correlations and anomaly propagation within the system. This analysis aims to understand the intrinsic relationships between data from different monitoring points and quantify the generation and propagation probabilities of abnormal signals. This analysis is based on a multi-sensor network deployed on the seat. For example, multiple triaxial vibration acceleration sensors are installed at key locations in the seat base and backrest frame, forming a vibration monitoring point array; a distributed pressure sensor array is embedded inside the seat cushion and backrest, forming a pressure monitoring point network; and displacement and current sensors are installed near actuators and dampers. These monitoring points synchronously collect data at a sampling rate of at least 1 kHz, providing a multi-dimensional, high-precision time-series signal source for analysis. All sensor data is transmitted to the central processing unit via CAN bus or Ethernet and timestamped to ensure data temporal consistency.

[0024] The first step in the analysis is to determine the time delay between signals from two physically related monitoring points. Based on the seat's mechanical structure model and signal propagation path, the system intelligently selects sensor pairs with direct connections or conduction relationships. For example, for vibration analysis, two accelerometers arranged along the main beam of the seat frame would be selected. The system calculates the cross-correlation function of the two sensor signals within a sliding time window. By finding the global maximum value of the cross-correlation function, the time delay of one signal relative to the other is accurately estimated. This time delay not only contains information about the propagation speed of stress waves in the metal structure but also implies potential information such as the condition of connectors and the degree of material fatigue. A time delay that deviates significantly from the historical normal range may indicate changes in structural stiffness, loose bolts, or microcracks in the material.

[0025] The system introduces a short-time analysis window of 100-500 milliseconds to segment the continuous sensor signal stream into a series of overlapping time segments. A Hanning window function is used to weight the signal within each window to reduce spectral leakage. For each monitoring point, within its short-time window, the system analyzes the amplitude distribution characteristics of the sensor signal, calculating its peak-to-peak value, root mean square value, and interquartile range. Then, the signals from downstream monitoring points are precisely aligned based on the calculated time delay values. After alignment, the system uses statistical distance metrics (such as KL divergence or JS divergence) to compare the amplitude distribution of the upstream monitoring point signal within a certain short-time window with the amplitude distribution of the downstream monitoring point signal within the corresponding time window. The degree of difference in their distribution characteristics is quantified into a dimensionless index called the signal fluctuation factor. This factor reflects the degree of distortion in the energy distribution characteristics of the signal as it propagates from one point to another. An abnormally high fluctuation factor may indicate the presence of nonlinear elements, impacts, or abnormal energy dissipation in the propagation path.

[0026] After obtaining the time-domain characteristics, the analysis delves further into the frequency domain. The system performs a 2048-point Fast Fourier Transform on the sensor signals within all short-time windows of adjacent monitoring points to obtain their respective power spectral density estimates. For each pair of aligned short-time window spectra, the system compares the amplitude and phase differences at each frequency point. Amplitude differences reveal abnormal attenuation or amplification of signal energy at different frequency components; phase differences reflect the wave velocity changes of different frequency components during signal propagation. These frequency-domain differences, combined with the previously calculated signal fluctuation factor, are input into a probabilistic calculation model based on Bayesian inference. This model comprehensively considers sensor accuracy, environmental noise levels, and historical normal fluctuation ranges to assess how likely the observed differences at a specific frequency are caused by abnormal interference (such as structural resonance, electromagnetic interference, component wear, or lubrication failure) rather than normal system noise or changes in operating conditions. The output is a frequency function, representing the interference probability values ​​of the sensor signals at each frequency point, with these probability values ​​continuously distributed between 0 and 1.

[0027] The system analyzes the interference probability at all frequency points using an adaptive threshold algorithm. This threshold is dynamically determined based on the overall shape and local features of the probability distribution. The system identifies specific frequency bands where the interference probability is consistently significantly higher than the background level; these are marked as interference frequencies. These interference frequencies are often associated with specific types of fault mechanisms; for example, a high interference probability at a certain characteristic frequency may indicate rotor imbalance, bearing defects, or abnormal gear meshing. The system also records characteristics such as bandwidth, center frequency, and probability amplitude for each interference frequency.

[0028] The analysis process needs to move from local features to a system-level evaluation. The system aggregates the interference frequency information obtained from the analysis of all adjacent monitoring point pairs to construct an interference frequency distribution map of the entire sensor network. The location, propagation path, energy attenuation law, and frequency domain characteristics of these interference frequencies within the sensor network are observed. Some faults may only generate narrowband interference frequencies locally, while others may trigger broadband, correlated interference frequency clusters at a series of monitoring points. Based on these distribution patterns, the system uses graph theory algorithms and statistical pattern recognition methods to calculate a comprehensive index—the nonlinear error probability. This probability value reflects the likelihood that the signal anomalies captured by the entire sensor network are caused by a systemic nonlinear fault (such as damper failure, actuator jamming, control system instability, or structural damage), rather than independent local disturbances or measurement noise.

[0029] The entire diagnostic target impact analysis process runs iteratively on a dedicated computing unit equipped with a digital signal processor, employing multi-threaded parallel processing to improve computational efficiency. Its output—the nonlinear error probability and its associated interference frequency distribution map—provides precise targets and directions for subsequent parameter optimization and fault diagnosis, transforming fault diagnosis from a simple threshold alarm into a deep analysis process with causal inference capabilities. The computational complexity of this analysis process is linearly related to the number of monitoring points and logarithmically related to the analysis frequency resolution. After optimization, all calculations can be completed within seconds, meeting the requirements of real-time diagnosis.

[0030] Example 3: The system first constructs a profile deviation vector, an n-dimensional mathematical structure where each component represents the standardized difference between the real-time profile and the retrospective profile on a specific feature parameter. These feature parameters cover dozens of specific features under multiple diagnostic targets, such as vibration, pressure, and response time. For example, vibration diagnostic targets may include features such as vibration energy across multiple frequency bands, principal mode frequencies, and damping ratios; pressure distribution diagnostic targets may include pressure values ​​in various areas of the seat, pressure gradients, and symmetry indices; response time diagnostics include step response time, settling time, and overshoot. All feature parameters are standardized before deviation calculation to eliminate dimensional influences. All these difference values ​​together constitute a multi-dimensional vector comprehensively describing the system's deviation from its normal state. The system extracts a set of target sensor data deviation magnitudes from this profile deviation vector. Here, "target sensor data" refers to sensor readings listed as key diagnostic parameters. The deviation magnitude is the absolute value of the deviation of each key parameter, used to quantify its degree of deviation. The set of all these magnitudes reflects the overall deviation of the key parameters. The system also calculates the confidence weight for each deviation modulus, which is determined based on the sensor’s historical reliability and current signal quality.

[0031] The system calculates the statistical variance of the set of deviation magnitudes. The magnitude of the variance characterizes whether the deviation behavior of these key parameters is consistent. A high variance indicates that some parameters deviate significantly, while others are close to normal; this inconsistent deviation pattern often suggests the type of fault. The system presets a dynamically adjusted variance threshold, which is derived from statistical analysis of a large amount of historical fault data and takes into account current operating conditions and environmental factors, to distinguish between random fluctuations and systemic faults.

[0032] When the variance of the set of deviation magnitudes is greater than or equal to the variance threshold, the system initiates the calculation of the failure probability. This calculation involves a comprehensive evaluation model, the core formula of which can be expressed as: , in: This represents the calculated failure probability, and its value ranges from 0 to 1. This represents the variance of the set of target sensor data deviation magnitudes obtained from the aforementioned calculations. This represents the preset variance threshold, which is a constant set based on system characteristics and historical experience. It is a weighting coefficient determined by the nonlinear error probability output from the diagnostic target impact analysis phase, along with other prior knowledge (such as sensor reliability, historical occurrence rate of this type of failure, etc.). Its role is to weight and adjust the importance of variance. The use of exponential functions and natural constants makes the output probability value... It can produce a non-linear response to changes in input parameters, which conforms to the actual laws of the probability of fault occurrence.

[0033] Calculated failure probability These constraints form the direct basis for establishing parameter optimization constraints. Essentially, parameter optimization constraints are a set of rules or boundary conditions used to limit the search space and optimization direction of subsequent parameter-oriented optimization algorithms. Constraint types include hard constraints and soft constraints: hard constraints strictly limit the physically feasible range of parameters, such as the current not exceeding the maximum allowable value of the coil; soft constraints, based on fault probability and diagnostic results, guide the search direction, such as prioritizing the optimization of damping parameters when the probability of abnormal vibration is high. The stringency of the constraints is also proportional to the fault probability value; the higher the probability, the tighter the constraints may be applied to accelerate the optimization speed and increase the likelihood of finding an effective solution. The system also establishes a constraint priority system to ensure that critical safety-related constraints receive the highest priority execution.

[0034] The parameter-oriented optimization process operates under established constraints, employing a combination of various optimization algorithms. Its output is a set of optimized system parameter values, representing the best or near-best system performance achievable by adjusting control parameters under the current fault state. The optimization process not only pursues optimal performance but also considers the smoothness and feasibility of parameter changes, avoiding secondary shocks caused by abrupt parameter jumps. This optimization result is ultimately used for fault diagnosis; fault type determination is a multi-feature fusion pattern matching and decision-making process. The system integrates the optimization results, the original deviation vector features, and the fault probability. The system compares the interference frequency features obtained during the diagnostic target impact analysis phase with records in the fault feature database. This database is continuously updated using machine learning methods, and each record is associated with a fault mode and its typical multidimensional features. By calculating the similarity measure of the multidimensional feature space and employing a random forest-based classification algorithm, the system identifies the best-matching fault type. The system also outputs a confidence score for the fault type, reflecting the reliability of the diagnostic results.

[0035] The assessment of fault severity relies more on quantitative analysis, including the magnitude of parameter adjustments in the optimization results, the residual deviation between the optimized system state and the ideal state, and the fault probability. The size of the fault itself and the degree of degradation of key features are crucial indicators for quantifying severity. The system employs fuzzy logic reasoning to map continuous, quantified indicators to discrete severity levels. For example, severity is divided into three levels: mild, moderate, and severe, each with a corresponding threshold range and recommended treatment measures. The assessment process also considers the fault's development trend, determining whether the fault is in a stable state, slowly degrading, or accelerating by comparing recent diagnostic results. Diagnostic conclusions are output in the form of fault type, fault severity, confidence score, and predicted development trend, providing clear and direct evidence for maintenance decisions.

[0036] Example 4: See Figure 4In the parameter-oriented optimization stage of the magnetorheological seat fault diagnosis method, the core objective is to find a set of operating parameters that optimize system performance under defined constraints through systematic search and evaluation. This process begins with establishing the control range of key diagnostic parameters. This range is not fixed but dynamically defined based on the seat's physical limits, safe operating range, real-time operating conditions, and optimization constraints determined by previous fault diagnosis analysis. For example, the lower limit of the control current of the magnetorheological damper may be determined by the current value that maintains the minimum damping effect, while the upper limit is limited by multiple constraints such as the allowable temperature rise of the coil, system energy consumption indicators, and power output capacity. For the actuator displacement parameters that adjust the seat posture, the range is jointly set by the physical stroke of the mechanical structure, the safety margin to avoid interference with surrounding components, and the current load conditions. The control range of each parameter is represented as a closed interval, accompanied by a recommended operating value.

[0037] Within a defined control interval, the system creates an initial solution set based on the current parameter set (typically parameter values ​​read under fault conditions). This initial solution set contains dozens of candidate parameter combinations, generated using Latin hypercube sampling within their respective control intervals. This ensures the solution set adequately covers the entire possible parameter space while maintaining sampling uniformity across parameter dimensions. Each candidate solution represents a complete set of system parameter settings. The size of the solution set is dynamically adjusted based on the dimensionality of the parameter space and real-time optimization requirements, typically ranging from 20 to 100 solutions.

[0038] Fitness evaluation is performed on each candidate solution in the initial solution set. The fitness function is designed to integrate multiple performance indicators, and its weights are dynamically adjusted based on the priority of the current diagnostic objective. The evaluation process is completed through a high-fidelity system simulation model or by deploying a lightweight test program directly on the seat control system. The simulation model employs multibody dynamics modeling and includes multiple subsystems such as the seat's mechanical structure, hydraulic system, control system, and human body model. The test program applies carefully designed test signals to the actual system to avoid interfering with normal operation. The simulation or test run simulates the seat's dynamic response under specific parameter combinations and collects key output performance data, such as the root mean square value of vibration acceleration, pressure distribution non-uniformity, stabilization time to the target posture, and overshoot. These output data are quantified into a comprehensive fitness score through a weighted aggregation function to evaluate the quality of the parameter set. A higher score indicates better overall performance of the parameter combination.

[0039] Based on the fitness evaluation results and the previously established parameter optimization constraints, the system determines the next optimization direction and step size. The optimization direction indicates which direction in the parameter space should be adjusted to improve fitness; this decision may be based on a comparison of the distribution of better and worse solutions in the solution set, or by estimating the performance gradient through a simple local regression model. The optimization step size determines the magnitude of each adjustment; a larger step size facilitates rapid exploration, while a smaller step size allows for fine-tuning near high-quality solutions. The step size is set with reference to the size of the parameter control interval, the dispersion of the current solution set, and the smoothness estimate of the fitness function.

[0040] Using a defined optimization direction and step size, the system begins iteratively updating the initial solution set. This typically employs a framework of an improved swarm intelligence optimization algorithm, where each candidate solution in the solution set is treated as an independent agent. Based on their own historical experience, the experience of other high-performing individuals in the swarm, and global optimization guidance, they purposefully move within the parameter space, thereby generating a new generation of solution sets. The iterative process employs an elitist retention strategy to ensure that high-performing solutions are not randomly corrupted. Simultaneously, a diversity preservation mechanism is introduced to prevent the population from prematurely converging to local optima.

[0041] During the optimization process, the system also establishes a parameter energy consumption evaluation function in parallel. This function estimates the power consumption of the seat system under each set of candidate parameters, which may include the copper loss of the damper coil, the drive power consumption of the actuator motor, and the standby power consumption of the control system. The energy consumption model is established based on the physical parameters and operating characteristics of each component and is verified through real-time current and voltage measurements. Simultaneously, based on the performance data output from simulations or tests, the system calculates another quality evaluation function, which directly reflects the degree of improvement of the diagnostic target, such as vibration suppression effect score or pressure regulation accuracy score. Energy consumption and quality often have a trade-off relationship. To find a balance point, the system establishes a balanced adaptive function, which combines the energy consumption score and the quality score in a weighted or compromised manner. The weighting coefficients are dynamically adjusted according to the operating mode (such as energy saving priority or performance priority), as shown in Table 1.

[0042] Table 1: Screening of Parameter Optimization Schemes

[0043] Based on the fitness function, the system screens optimal solutions generated during all iterations, as well as previously retained excellent solutions. For example, the system calculates and compares the overall performance of various parameter combinations. The screening process considers not only the absolute value of the fitness value but also the diversity of solutions, the degree of constraint satisfaction, and the smoothness of parameter changes. Ultimately, the optimal solution with the highest fitness value, satisfying all constraints, and exhibiting reasonable parameter changes is output as the result of parameter-oriented optimization. This set of parameters represents the operational settings that achieve the best balance between system performance and energy consumption under the current fault state, providing the most direct decision-making basis for subsequent fault diagnosis conclusions and directly guiding the system's fault-tolerant operation or maintenance adjustments. The entire optimization process is completed on an embedded computing platform or edge server, employing parallel computing and algorithm optimization techniques. Its computational efficiency is guaranteed through algorithm design and hardware acceleration to meet the response speed requirements of practical applications. The optimization process can typically be completed within minutes, with the specific time depending on the dimensionality of the parameter space and the required optimization accuracy.

[0044] Example 5: In the parameter-oriented optimization process of magnetorheological seat fault diagnosis, the effectiveness of the iterative update strategy directly determines the accuracy and efficiency of the optimization results. The core of this implementation lies in the refined management and dynamic adjustment of the iterative process of each candidate solution in the solution set, rather than using fixed update rules. The system establishes an independent iterative trajectory for each candidate solution participating in the iteration. This trajectory records the parameter value sequence of the solution in each iteration process and its corresponding fitness evaluation results. After each iteration, the system identifies the trajectory based on the latest fitness value of the solution. This identification can be understood as labeling the trajectory with a tag reflecting its recent performance change trend, such as "continuous improvement", "fluctuation and stagnation" or "continuous degradation".

[0045] To conduct phased evaluations of the iterative process, the system configures iterative evaluation intervals, typically in units of a fixed number of iterations, such as ten iterations per interval. At the end of each evaluation interval, the system comprehensively analyzes the iterative trajectories of all solutions. The analysis focuses on identifying the update status of each solution within that interval, specifically comparing the changes in fitness values ​​between the start and end points of the interval and observing fluctuations during these changes. Based on this analysis, the system refines the solution classification from a simple "good" and "bad" to three more instructive evaluation categories: excellent solution evaluation, exploratory evaluation, and inferior solution evaluation. Solutions belonging to the excellent solution evaluation category exhibit a stable and significant increase in fitness values ​​within the evaluation interval, indicating potential for further exploration in their current parameter region. Solutions in the exploratory evaluation category may not show a significant increase in fitness values ​​or may even fluctuate slightly, but their parameter positions are in unknown regions, and their exploratory behavior may bring new diversity to the population. The inferior solution evaluation category includes solutions whose fitness values ​​significantly decrease within the evaluation interval, indicating that their update direction has been proven ineffective or harmful. The system no longer adopts a uniform update strategy for all solutions, but instead implements differentiated management based on their evaluation classification.

[0046] For solutions in the excellent solution evaluation category, the system configures a local surrogate model. This model utilizes the solution's recent iteration trajectory data and historical data from other excellent solutions in its neighborhood to construct a simplified mathematical model for predicting the effect of parameter fine-tuning. By analyzing this local surrogate model, the system can predict which fine-tuning direction is most likely to bring further performance improvement near the current excellent parameter point, thus generating a first reference optimization direction. This direction guides the refined local development of this type of solution. For solutions in the poor solution evaluation category, the system configures a penalized optimization identification layer. This layer analyzes the update directions adopted by these solutions within the most recent evaluation interval and identifies them as erroneous directions that need to be avoided. These identified erroneous directions are recorded in a list called "Window Improvement Taboos." In subsequent iterations, the system will try to prevent other solutions from using these proven ineffective directions for updates, thereby reducing invalid searches and improving efficiency.

[0047] The specific execution of search self-optimization management integrates the above information. For solutions within the excellent solution evaluation category, the system primarily performs fine-tuning and iterative updates based on the first reference optimization direction, while simultaneously referencing the window improvement taboo list to avoid repeating past mistakes. The update step size is typically set small to allow for refined searching. For solutions within the exploratory evaluation category, the system employs a hybrid exploration strategy. Its update direction combines the first reference optimization direction with a certain degree of random perturbation, aiming to maintain the ability to explore unknown regions while providing some guidance. The step size is set more flexibly. For solutions within the inferior solution evaluation category, the system configures a strong random factor to perform its iterative updates, allowing it to escape potentially low-quality regions with greater randomness, reintroducing population diversity. Its update direction is strictly constrained by the window improvement taboo list.

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fault diagnosis method for a magnetorheological seat, characterized in that, The method includes: Acquire the current state data of the magnetorheological seat and establish a fault feature database that maps to the current state data; Diagnostic targets are configured according to the fault feature database, including vibration anomaly diagnosis, pressure distribution diagnosis, and response time diagnosis. Determine the target classification identifiers in the diagnostic objectives, wherein the target classification identifiers include primary diagnostic objectives, auxiliary diagnostic objectives, and reference objectives; After selecting key diagnostic parameters using the primary and secondary diagnostic objectives, an impact analysis of the diagnostic objectives on the key diagnostic parameters is performed. After establishing parameter optimization constraints based on the results of the diagnostic target impact analysis, parameter-oriented optimization is performed, and the magnetorheological seat fault diagnosis is completed using the results of parameter-oriented optimization.

2. The magnetorheological seat fault diagnosis method as described in claim 1, characterized in that, The acquisition of the current state data of the magnetorheological seat includes: When the preset diagnostic cycle is met, a retrospective profile of the acquisition time of the target sensor data of the magnetorheological seat is obtained, wherein the retrospective profile includes a first normal state value and a first control parameter; A real-time profile of the magnetorheological seat is obtained, wherein the real-time profile includes a second current state value and a second control parameter.

3. The magnetorheological seat fault diagnosis method as described in claim 2, characterized in that, The retrospective profile of the acquisition time of the target sensor data of the magnetorheological seat includes: Multiple preset control parameters are used to obtain the target sensor data; With each preset control parameter as a constraint, the alienation state value of the target sensor data is statistically analyzed based on the historical sample library; Extract 0.5 times the minimum value of all alienated state values ​​and set it as the preset diagnostic cycle.

4. The magnetorheological seat fault diagnosis method as described in claim 1, characterized in that, The impact analysis of the diagnostic objectives of executing key diagnostic parameters includes: The time delay between the signals from the two monitoring points is determined based on the correlation between the sensor signals at the two monitoring points. The signal fluctuation factor is determined based on the amplitude distribution of the sensor signal at the monitoring point within a short time window and the amplitude distribution of the sensor signal at the next monitoring point after a time delay within the short time window. Based on the amplitude difference, phase difference, and signal fluctuation factor of the sensor signals contained in all short-time windows of adjacent monitoring points in the frequency domain, the interference probability of the sensor signals at each frequency of the monitoring points is determined. The interference frequency is obtained based on the interference probability of the sensor signals at the monitoring point at all frequencies; The nonlinear error probability is obtained based on the distribution of interference frequencies at all adjacent monitoring points.

5. The magnetorheological seat fault diagnosis method as described in claim 1, characterized in that, The establishment of parameter optimization constraints based on the impact analysis results of the diagnostic objectives includes: Retrieve the set of target sensor data deviation magnitude values ​​that satisfy the profile deviation vector. If the variance of the set of target sensor data deviation magnitude values ​​is greater than or equal to the variance threshold, calculate the failure probability. Parameter optimization constraints are established based on the aforementioned failure probabilities.

6. The magnetorheological seat fault diagnosis method as described in claim 1, characterized in that, The execution parameter-oriented optimization includes: After establishing the control range of the parameters, an initial solution set is created based on the current parameter set; After performing fitness evaluation of solutions within the initial solution set, the optimization direction and optimization step size are established through parameter optimization constraints and fitness evaluation results; The initial solution set is iteratively updated using the optimization direction and the optimization step size. The parameter optimization is completed based on the iterative update results.

7. The magnetorheological seat fault diagnosis method as described in claim 6, characterized in that, The step of iteratively updating the initial solution set using the optimization direction and the optimization step size includes: For each solution, an iterative trajectory is established, and the iterative trajectory is identified by the fitness value of the solution in each iteration. Configure an iterative evaluation interval, identify the update status of the iterative trajectory within the iterative evaluation interval, and generate an evaluation classification, which includes a good solution evaluation classification, an exploration evaluation classification, and a poor solution evaluation classification. Search self-optimization management is performed through iterative updates based on the evaluation classification.

8. The magnetorheological seat fault diagnosis method as described in claim 7, characterized in that, The search self-optimization management that iteratively updates based on the evaluation classification includes: In the optimal solution evaluation classification configuration, a local proxy model is used to predict the improvement trend and generate a first reference optimization direction. In the inferior solution evaluation classification configuration, a penalty optimization identification layer is configured, and the penalty optimization identification layer is used to identify the direction of error improvement and establish window improvement taboos; The solution fine-tuning iteration update within the optimal solution evaluation category is performed using the first reference optimization direction and window improvement taboo, and the hybrid exploration iteration update within the exploration evaluation category is performed using the first reference optimization direction and window improvement taboo. A random factor is configured to perform the iteration update of the solution within the inferior solution evaluation category.

9. The magnetorheological seat fault diagnosis method as described in claim 6, characterized in that, The execution parameter-oriented optimization also includes: Establish an energy consumption evaluation function for the parameters; Establish a balance adaptation function based on the energy consumption evaluation function and the quality evaluation function; Based on the aforementioned equilibrium fitness function, the optimization scheme is screened for parameter-oriented optimization, and the optimization scheme screening result is output as the parameter-oriented optimization result.

10. The magnetorheological seat fault diagnosis method as described in claim 1, characterized in that, The process of using parameter-oriented optimization results to diagnose magnetorheological seats includes determining the fault type and severity based on the parameter-oriented optimization results.

Images