Seat adjustment parameter dynamic optimization method for commercial vehicle

Abstract

The invention belongs to the field of seat control, and relates to a dynamic optimization method for seat adjustment parameters of a commercial vehicle. The method comprises the following steps: collecting and processing a motor signal to obtain a current vector sequence in an angle domain; extracting a composite feature matrix capable of reflecting the current state of the seat system through time-frequency analysis; taking the composite feature matrix sequence as an observation value of a factorial hidden Markov model, and diagnosing and distinguishing posterior probabilities of parallel hidden state chains corresponding to passenger loads and mechanical resistance; when the deviation between the diagnosed system state and the ideal state is too large, a multi-objective optimization program is started, and a plurality of performance indexes such as the adjustment time, the driving torque fluctuation rate and the predicted temperature rise are subjected to collaborative optimization through a non-dominated sorting genetic algorithm; and the optimal motor control parameters of the next period, which can balance the adjustment time, the riding comfort and the energy consumption, can be efficiently solved. The comprehensive requirements of a modern commercial vehicle for efficient seat adjustment, comfortable experience and reliable operation can be met.

Description

Technical Field

[0001] This invention belongs to the field of seat control, and in particular relates to a method for dynamically optimizing seat adjustment parameters in commercial vehicles. Background Technology

[0002] In the process of upgrading the driving and riding experience of modern commercial vehicles, electrically adjustable seats have transformed from optional configurations to core standard components. Driven by a micro DC motor, and in conjunction with gearbox reduction and lead screw transmission mechanisms, they can achieve multi-degree-of-freedom adjustments such as seat fore-and-aft movement, backrest angle, and lumbar support, meeting the comfort needs of passengers of different weights and in different riding scenarios. Current mainstream seat adjustment control systems complete the calibration of core control parameters such as motor speed, acceleration threshold, and current protection parameters during the product development stage, and maintain these parameters fixed throughout their entire lifecycle. While this design approach has the advantages of simple structure and low hardware cost, and can meet basic adjustment functions, it lacks the ability to sense and adaptively adjust to real-time operating conditions.

[0003] From a practical application perspective, the load state of a seating system is constantly changing: on the one hand, the external load is affected by passenger weight and posture habits, exhibiting non-fixed characteristics; on the other hand, the internal mechanical resistance changes continuously over the usage period. For example, the friction coefficient increases due to aging gearbox lubricating oil, the transmission clearance increases due to long-term wear of the lead screw, and the risk of jamming caused by thermal expansion and contraction of components in low-temperature environments, all of which can cause the mechanical resistance to deviate from the initial design value. Fixed control parameters cannot respond to the above dynamic changes, directly leading to two typical problems: in light-load scenarios, because the parameters are conservatively set according to heavy-load conditions, the motor operates in an inefficient range, resulting in slow adjustment response and poor user experience; in heavy-load or high-resistance scenarios, fixed parameters are difficult to provide sufficient driving torque, easily causing motor overload, obvious shaking and abnormal noise during adjustment, and in severe cases, even causing transmission mechanism jamming, shortening the service life of the seating system. Further analysis of the load monitoring mechanism of existing control systems reveals that current solutions generally only judge the load size by collecting the total motor current, which cannot effectively distinguish between passenger load and mechanical resistance, two different types of load sources, thus limiting the health status assessment and fault warning capabilities of the control system.

[0004] To accommodate various potential operating conditions, compromise or conservative parameter settings are often adopted. This causes the motor to deviate from its optimal operating point most of the time, resulting in unnecessary energy consumption and heat accumulation, shortening the lifespan of the motor and related electronic components, and increasing later maintenance costs. Therefore, it can be seen that existing electric seat adjustment control systems for commercial vehicles, due to issues such as fixed parameters, insufficient load recognition capabilities, and a single monitoring mechanism, can no longer meet the comprehensive needs of modern commercial vehicles for efficient adjustment, comfortable experience, and reliable operation. Summary of the Invention

[0005] Therefore, the purpose of this invention is to propose a dynamic optimization method for seat adjustment parameters in commercial vehicles, in order to solve the technical problems of traditional electric seats in commercial vehicles, which, due to their fixed control parameters, cannot adapt to changing passenger loads and mechanical resistance, resulting in poor adjustment experience, high energy consumption, and shortened lifespan.

[0006] To address the above problems, the present invention provides a technical solution for a dynamic optimization method of seat adjustment parameters in a commercial vehicle: A method for dynamically optimizing seat adjustment parameters in a commercial vehicle includes the following steps: S1. Acquire the three-phase electrical signal and rotor position signal of the motor, and obtain the current vector sequence that can eliminate speed interference in the angle domain through coordinate transformation and equal angle resampling; perform time-frequency analysis on the current vector sequence, and extract the time-frequency entropy and fractal dimension in a specific frequency band related to the physical characteristics of the seat transmission chain, and construct a composite feature matrix that can quantitatively characterize the current operating state of the system. S2. Use the composite feature matrix as the observation value of the factorial hidden Markov model, and use the observation value to estimate the posterior probability of the parallel hidden state chain corresponding to passenger load and mechanical resistance. S3. The adjustment time, driving torque fluctuation rate and predicted temperature rise are taken as multi-objective optimization objectives, and the current state distribution is calculated based on the posterior probability. If the divergence between the current state distribution and the preset nominal distribution exceeds the preset threshold, the initial population search domain of the non-dominated sorting genetic algorithm is defined by the posterior probability, and the multi-objective optimization problem is solved to obtain the target motor speed, acceleration and current loop feedforward compensation coefficients for the next control cycle.

[0007] Furthermore, in step S1, the method for obtaining the current vector sequence in the angle domain includes: The collected three-phase stator currents are transformed by Clark to obtain the current components in a two-phase stationary coordinate system. By combining the rotor position signal, the current components are transformed to the dq synchronous rotating coordinate system through Park transformation to obtain the quadrature axis current and the direct axis current. For the dq axis current and voltage signals, linear interpolation is performed based on the rotor position signal, and resampling is performed at preset fixed angle intervals to obtain the current vector sequence in the angle domain.

[0008] Preferably, the linear interpolation based on the rotor position signal is achieved by using a cubic spline interpolation algorithm.

[0009] Further, in step S1, the method for obtaining the composite feature matrix includes: The cross-axis current sequence in the angular domain is processed using short-time Fourier transform, and a time-frequency spectrum is generated. Based on the gear meshing frequency and bearing natural frequency of the seat drive chain, the time spectrum is divided into three frequency bands: the load change band, the gear meshing band, and the bearing friction band, which are respectively the frequency bands that increase sequentially. The Shannon time-frequency entropy of the energy distribution in each frequency band is calculated, and the fractal dimension of the signal in each frequency band is calculated by the box-counting dimension method. The six eigenvalues, including the Shannon time-frequency entropy and the fractal dimension of each frequency band, are combined into a composite feature. When the seat motor is running continuously, the composite feature vectors are continuously generated, and these continuous composite feature vectors are arranged in chronological order to form a composite feature matrix.

[0010] Preferably, the frequency range of the load variation belt is 0-5Hz, the frequency range of the gear meshing belt is 50-150Hz, and the frequency range of the bearing friction belt is 200-500Hz.

[0011] Furthermore, the method for calculating the posterior probability includes: The implicit state chain of passenger load includes three states: no load, standard load, and overload; the implicit state chain of mechanical resistance includes two states: normal resistance and increased resistance. Using a forward-backward algorithm, based on the observed composite feature matrix, the posterior probability of each possible state in the two parallel hidden state chains of passenger load and mechanical resistance is iteratively calculated at the current moment.

[0012] Furthermore, in step S3, the adjustment time is the total time it takes for the seat to move from the starting position to the target position. ; The driving torque fluctuation rate is the standard deviation of the quadrature axis current sequence during the adjustment process; The predicted temperature rise is based on the motor winding resistance. Heat capacity and the root mean square current during the adjustment process Calculated temperature rise Temperature rise The calculation formula is: .

[0013] Further, in step S3, the current state distribution is the current joint probability distribution of the seat system obtained by multiplying the posterior probabilities of passenger load and mechanical resistance; the nominal distribution is the joint probability distribution of the seat system under ideal conditions; wherein, the probability corresponding to the joint state of standard load and normal resistance is 1, and the probability corresponding to all other joint states is 0; the divergence is the degree of deviation between the current joint probability distribution of the seat system and the joint probability distribution under ideal conditions.

[0014] Further, in step S3, the method for defining the initial population search domain of the non-dominated sorting genetic algorithm using the posterior probability includes: If the posterior probability that the passenger load is in an overloaded state is greater than the first confidence threshold, then the upper limit of the initial search range for the target speed and acceleration of the motor is reduced, and the upper limit of the initial search range for the current loop feedforward compensation coefficient is increased. If the posterior probability of the mechanical resistance being in an increasing resistance state is greater than the second confidence threshold, then the initial search lower bound of the current loop feedforward compensation coefficient is increased.

[0015] Preferably, the first confidence threshold is set to 0.8, and the second confidence threshold is set to 0.7; If the posterior probability that the passenger load is in an overloaded state is greater than 0.8, then the lower limit of the initial search range of the motor target speed will be reduced by 20%, the upper limit of the initial search range of acceleration will be reduced by 15%, and the upper limit of the initial search range of the current loop feedforward compensation coefficient will be increased by 30%. If the posterior probability of the mechanical resistance being in an increasing resistance state is greater than 0.7, then the lower limit of the initial search range of the current loop feedforward compensation coefficient is increased by 25%.

[0016] The beneficial effects of this invention are as follows: This invention employs a factorial hidden Markov model for deep analysis of motor signals, effectively separating and identifying two different load sources: passenger load and transmission chain mechanical resistance. This overcomes the limitation of existing technologies that can only monitor total current but cannot distinguish load sources. Furthermore, this invention uses a non-dominated sorting genetic algorithm to collaboratively optimize multiple performance indicators such as settling time, drive torque fluctuation rate, and predicted temperature rise, ultimately obtaining a set of motor control parameters that achieve a better balance between efficiency, smoothness, and energy consumption. Without increasing hardware costs, this improves passenger comfort during seat adjustment, reduces energy consumption and heat in the drive system, and helps extend the mechanical life of the transmission mechanism. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of a method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] A specific embodiment of the method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to the present invention: like Figure 1 As shown, a method for dynamically optimizing seat adjustment parameters in a commercial vehicle includes the following steps: S1. Acquire the three-phase electrical signals and rotor position signals of the motor, and obtain the current vector sequence that can eliminate speed interference in the angle domain through coordinate transformation and equal-angle resampling; perform time-frequency analysis on the current vector sequence, and extract the time-frequency entropy and fractal dimension in a specific frequency band related to the physical characteristics of the seat transmission chain, and construct a composite feature matrix that can reflect the current operating state of the seat system.

[0020] In this step, the current sensor built into the motor driver can acquire the three-phase stator current signal, while the rotor position signal is acquired through an incremental encoder mounted coaxially with the motor. Differential calculations are then performed on the rotor position signal to obtain the real-time rotor speed. The three-phase ABC currents in the three-phase stationary coordinate system are converted into αβ currents in the two-phase stationary coordinate system. Then, using the rotor position signal, the αβ currents are converted into dq-axis currents in the synchronous rotating coordinate system. To eliminate the influence of speed fluctuations on signal analysis, based on the acquired rotor position sequence, a cubic spline interpolation algorithm is used to resample the non-uniformly sampled dq-axis current and voltage signals in the time domain to an angle domain sequence with 0.1-degree intervals, obtaining analysis data that is strictly synchronized with the rotor position.

[0021] The d-axis and q-axis current sequences in the angular domain are decomposed into 32 uniformly distributed sub-frequency bands. Based on the design parameters of the seat transmission mechanism, such as the number of teeth on the gearbox gears and the lead screw, combined with the current motor speed, the theoretical frequency band ranges of the gear meshing frequency, lead screw rotation frequency, and their harmonic frequencies are calculated. The decomposed frequency bands are matched with the theoretical characteristic frequency ranges, and eight characteristic frequency bands containing key transmission information are selected. The time-frequency entropy of the signal energy distribution within each characteristic frequency band is calculated to obtain the degree of signal energy concentration; the fractal dimension of the frequency band signal is calculated using the box counting method, which represents the waveform irregularity. The 32 feature values ​​extracted from the eight characteristic frequency bands of the d-axis and q-axis currents are arranged in a predetermined order to obtain a composite feature vector within an observation time window. Continuous composite feature vectors constitute a composite feature matrix.

[0022] To convert a time-domain signal into an angle-domain signal that is insensitive to changes in motor speed, in a preferred embodiment, the method for obtaining the current vector sequence in the angle domain includes: The collected three-phase stator currents (ia, ib, ic) are transformed by Clark to obtain the current components (iα, iβ) in the two-phase stationary coordinate system. By combining the rotor position signal θ, the current components (iα, iβ) are transformed to the dq synchronous rotating coordinate system through Park transformation to obtain the quadrature axis current iq and the direct axis current id; For the dq axis current signal, linear interpolation is performed based on the rotor position signal, and resampling is performed at fixed angle intervals of 0.1 degrees to obtain the current vector sequence in the angle domain.

[0023] For example, at a certain moment, the collected three-phase currents ia, ib, and ic are 5 amps, -2.5 amps, and -2.5 amps, respectively. After Clark transformation, the current components iα in the stationary coordinate system are found to be 5 amps and iβ to be 4.33 amps. If the rotor position signal θ is 30 degrees, the quadrature-axis current iq in the synchronous rotating coordinate system is calculated to be 4.97 amps using Park transformation. This value is proportional to the motor output torque, while the direct-axis current id is 0.5 amps, which is related to excitation. To eliminate the influence of speed fluctuations during motor adjustment on signal analysis, equal-angle resampling is further performed. The original data is collected at fixed time intervals. When the motor is running at high speed, the rotor rotation angle between adjacent sampling points is large, while it is small at low speed. By performing linear interpolation based on the rotor position signal, a data point can be generated at every 0.1-degree rotor rotation angle. For example, if there are two original sampling points at rotor positions of 15.23 degrees and 15.35 degrees, the corresponding iq and id values ​​at the 15.30-degree position can be calculated through interpolation. Therefore, regardless of the motor speed, the result is a sequence of 3600 evenly distributed data points per revolution.

[0024] In a preferred embodiment, the method for obtaining the composite feature matrix includes: The cross-axis current iq sequence in the angular domain is processed using short-time Fourier transform to generate a time-frequency spectrum. Based on the gear meshing frequency and bearing natural frequency of the seat drive chain, the time spectrum is divided into three frequency bands: the load variation band of 0-5Hz, the gear meshing band of 50-150Hz, and the bearing friction band of 200-500Hz. The Shannon time-frequency entropy of the energy distribution in each frequency band is calculated, and the fractal dimension of the signal in each frequency band is calculated by the box-counting dimension method. The six eigenvalues, including the Shannon time-frequency entropy and the fractal dimension of each frequency band, are combined into a composite eigenvector. When the seat motor is running continuously, the composite feature vectors are continuously generated, and these continuous composite feature vectors are arranged in chronological order to form a composite feature matrix.

[0025] For example, applying a short-time Fourier transform with a window function length of 256 data points and an overlap rate of 50% to the cross-axis current iq sequence in the angle domain yields a time-frequency spectrum, representing the variation of signal frequency components with seat position or angle. The time-frequency spectrum can be divided into several bands. For instance, the low-frequency band (0-5Hz) is mainly caused by slow load fluctuations due to passenger weight or posture changes; the mid-frequency band (50-150Hz) corresponds to the vibration frequency generated when gears mesh in a gearbox, such as a 20-tooth gear rotating at 4 revolutions per second producing an 80Hz fundamental frequency; and the high-frequency band (200-500Hz) is related to impact vibrations caused by defects in the balls or rollers of rolling bearings.

[0026] Within each defined frequency band, Shannon time-frequency entropy represents the uniformity or disorder of energy distribution. Healthy gears have distinct meshing frequencies, concentrated energy, and low time-frequency entropy values, such as 1.5; while worn gears lead to increased harmonics and dispersed energy, resulting in higher time-frequency entropy values, such as 2.8. Fractal dimension, represented by the box-counting method, indicates the complexity and irregularity of signal waveforms. Signals from normally lubricated bearings are relatively smooth, with fractal dimension values ​​close to 1.2, while signals from dry friction or damaged bearings are coarser and more complex, potentially increasing to 1.7. The six values—time-frequency entropy and fractal dimension—for each of the three frequency bands (e.g., 1.5, 1.2, 2.8, 1.4, 1.9, 1.7) are combined to form a composite feature vector. As the seat motor continues to run, these composite feature vectors are continuously generated. Arranging these continuous composite feature vectors in chronological order constitutes a composite feature matrix.

[0027] S2. Use the composite feature matrix as the observation value of the factorial hidden Markov model, and use the observation value to estimate the posterior probability of the parallel hidden state chain corresponding to passenger load and mechanical resistance.

[0028] Specifically, in this step, the factorial hidden Markov model is an improvement on the hidden Markov model, composed of multiple parallel and independent hidden state chains. In this embodiment, a factorial hidden Markov model containing two parallel hidden state chains is constructed. The first hidden state chain represents passenger load, and the second represents mechanical resistance. The composite feature matrix obtained in S1 is used as the observation sequence input to the factorial hidden Markov model. The expectation-maximization algorithm is used to train the model on a large amount of pre-collected sample data to obtain parameters such as the state transition probability matrix and observation probability matrix. In actual implementation, an approximate inference algorithm, such as the variational Bayesian inference method based on Gibbs sampling, is used to calculate the posterior probability that the two hidden state chains are simultaneously in any combination of states at the current time, based on the input real-time observation features.

[0029] In an optional embodiment, the posterior probability is calculated by: The implicit state chain of passenger load includes three states: no load, standard load, and overload; the implicit state chain of mechanical resistance includes two states: normal resistance and increased resistance. Using a forward-backward algorithm, based on the observed composite feature matrix, the posterior probability of each possible state in the two parallel hidden state chains of passenger load and mechanical resistance is iteratively calculated at the current moment.

[0030] The observed feature changes may be caused by changes in passenger load, mechanical resistance, or a combination of both. The factorial hidden Markov model contains two parallel and independent hidden state chains. For example, the first state chain represents passenger load, which can be empty, standard load, or overloaded, where empty corresponds to 0 kg, standard load to 75 kg, and overloaded to 120 kg. The second state chain represents mechanical resistance, which can be normal resistance or increased resistance due to factors such as rail jamming. Each combination of load and resistance states, such as standard load and increased resistance, corresponds to a probability distribution of a six-dimensional composite feature matrix. When processing the observed composite feature matrix using the forward-backward algorithm, the output might be: the probability of an overloaded passenger load state is 0.9, the probability of a standard load state is 0.08, and the probability of an empty passenger load state is 0.02; simultaneously, the probability of a normal mechanical resistance state is 0.7, and the probability of increased resistance is 0.3.

[0031] S3. The adjustment time, driving torque fluctuation rate and predicted temperature rise are taken as multi-objective optimization objectives, and the current state distribution is calculated based on the posterior probability. If the divergence between the current state distribution and the preset nominal distribution exceeds the preset threshold, the initial population search domain of the non-dominated sorting genetic algorithm is defined by the posterior probability, and the multi-objective optimization problem is solved to obtain the target motor speed, acceleration and current loop feedforward compensation coefficients for the next control cycle.

[0032] The core objective of this step is to automatically and intelligently adjust the motor control strategy based on the diagnosed system state to achieve optimal overall performance. First, three optimization objectives are defined to ensure the control goals of the seat system are comprehensive, taking into account user experience, efficiency, and lifespan. Simultaneously, optimization trigger conditions are set to avoid unnecessary and computationally intensive optimizations when the system is in a normal state. Furthermore, to improve optimization efficiency, the initial parameter search range of the genetic algorithm is dynamically adjusted using the posterior probability obtained in step S2 before starting the algorithm, accelerating its convergence speed. After running the non-dominated sorting genetic algorithm, a set of balanced solutions, i.e., the Pareto optimal solution set, is found among multiple objectives. One of these solutions is selected as the final output, ultimately producing a specific set of customized motor control parameters to be used in the next control cycle.

[0033] In a preferred embodiment, the adjustment time is the total time it takes for the seat to move from the starting position to the target position. The aforementioned drive torque fluctuation rate is the standard deviation of the quadrature-axis current sequence during the adjustment process; the aforementioned predicted temperature rise is based on the motor winding resistance. Heat capacity and the root mean square current during the adjustment process Calculated temperature rise Temperature rise The calculation formula is: .

[0034] For example, the first objective is the settling time, moving the seat from the rearmost to the frontmost position. The goal might be to optimize the total time t from the standard 8 seconds to 6 seconds. The second objective is the drive torque variability, which relates to the passenger's riding experience. The smoothness of the drive torque is measured by calculating the standard deviation of the quadrature-axis current iq sequence throughout the settling process. A smaller standard deviation, such as 0.2 amps, indicates smooth motor output torque and a smooth, vibration-free seat movement; while a larger standard deviation, such as 1.0 amps, indicates noticeable jerking during movement. The third objective is predicting temperature rise, which is crucial for ensuring the long-term reliability and safety of the motor. A simplified thermal model is used to estimate the impact of the settling process on the motor temperature. Assuming the motor winding resistance... It has a resistance of 0.8 ohms and a heat capacity of 0.8 ohms. The value is 150 joules per degree Celsius. If an adjustment process takes 6 seconds, the root mean square current during that period is... If the current is 5 amps, then the predicted temperature rise is... The temperature rise is 0.8 degrees Celsius. The goal of the optimization algorithm is to find a set of control parameters that, while meeting the requirements for settling time and comfort, predict the temperature rise as low as possible, thereby preventing the motor from overheating due to repeated or high-load adjustments and extending its service life.

[0035] In a preferred embodiment, the current state distribution is the joint probability distribution of the seat system obtained by multiplying the posterior probabilities of passenger load and mechanical resistance; the nominal distribution is the joint probability distribution of the seat system under ideal conditions. The probability corresponding to the joint state of standard load and normal resistance is 1, and the probability corresponding to all other joint states is 0; the divergence is the degree of deviation between the current joint probability distribution of the seat system and the joint probability distribution under ideal conditions.

[0036] For example, the two independent posterior probability distributions output by the factorial hidden Markov model are fused. For instance, if the estimated probability of a passenger load being at standard load is 0.9, the probability of overload is 0.1, and the probability of mechanical resistance being at normal resistance is 0.8, and the probability of increased resistance is 0.2, then the joint probability of the state of standard load and normal resistance is 0.72. In step S3, a joint probability distribution covering all six possible combinations can be obtained. The real-time joint probability distribution is compared with a preset ideal or nominal distribution, defined as the state of standard load and normal resistance with a probability of 1, and the probabilities of all other five combinations being 0. If the current system state is close to the ideal state, for example, the joint probability of standard load and normal resistance is 0.95, then the divergence value will be small, such as 0.05. However, if the system detects an overload state, its joint probability distribution will be significantly different from the nominal distribution, and the calculated divergence value may increase to 1.2. When this divergence value exceeds a preset threshold such as 0.5, the current operating condition has seriously deviated from the normal range, the existing control parameters are no longer applicable, and the multi-objective optimization program is started.

[0037] In a preferred embodiment, the method for defining the initial population search domain of a non-dominated sorting genetic algorithm using the posterior probability includes: If the posterior probability that the passenger load is in an overloaded state is greater than the first confidence threshold, then the upper limit of the initial search range for the target speed and acceleration of the motor is reduced, and the upper limit of the initial search range for the current loop feedforward compensation coefficient is increased. If the posterior probability of the mechanical resistance being in an increasing resistance state is greater than the second confidence threshold, then the initial search lower bound of the current loop feedforward compensation coefficient is increased.

[0038] For example, the first confidence threshold is set to 0.8, and the second confidence threshold is set to 0.7. Genetic algorithms typically generate an initial population randomly within a preset parameter space. In this embodiment, this parameter space is adjusted based on the state estimation results. For instance, the default search range for the target motor speed in the control parameters is 60 to 120 revolutions per minute. When the probability of detecting an overloaded passenger is as high as 0.85, exceeding the 0.8 threshold, a smoother, higher-torque start is required. The lower limit of the speed search range is reduced by 20%, becoming 48 to 120 revolutions per minute. Simultaneously, to avoid the impact under overload, the upper limit of the acceleration search is also reduced by 15% from 500 to 425. To provide sufficient driving torque, the upper limit of the current loop feedforward compensation coefficient search is increased by 30% from 1.2 to 1.56. Similarly, when the system detects an increase in mechanical resistance with a probability of 0.78, exceeding the 0.7 threshold, it indicates that greater force is needed to overcome friction. At this point, the search range of the current loop feedforward compensation coefficient is adjusted. The default search range of the current loop feedforward compensation coefficient is 0.6 to 1.2. The lower limit is increased by 25%, that is, from 0.6 to 0.75. This ensures that the initial population of the genetic algorithm contains individuals that can provide stronger instantaneous torque, avoids wasting computational resources in the ineffective low compensation region, and thus converges to the optimal control parameter set that can cope with high resistance conditions more quickly.

[0039] This invention can improve efficiency under light loads and ensure smoothness under heavy loads when adjusting the seats of commercial vehicles, while reducing energy waste and extending the life of motors and transmission components, and requires no additional hardware, thus reducing costs.

[0040] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for dynamically optimizing seat adjustment parameters in a commercial vehicle, characterized in that, Includes the following steps: S1. Acquire the three-phase electrical signal and rotor position signal of the motor, and obtain the current vector sequence that can eliminate speed interference in the angle domain through coordinate transformation and equal angle resampling; perform time-frequency analysis on the current vector sequence, and extract the time-frequency entropy and fractal dimension in a specific frequency band related to the physical characteristics of the seat transmission chain to construct a composite feature matrix that can reflect the current operating state of the seat system. S2. Use the composite feature matrix as the observation value of the factorial hidden Markov model, and use the observation value to estimate the posterior probability of the parallel hidden state chain corresponding to passenger load and mechanical resistance. S3. The adjustment time, driving torque fluctuation rate and predicted temperature rise are taken as multi-objective optimization objectives, and the current state distribution is calculated based on the posterior probability. If the divergence between the current state distribution and the preset nominal distribution exceeds the preset threshold, the initial population search domain of the non-dominated sorting genetic algorithm is defined by the posterior probability, and the multi-objective optimization problem is solved to obtain the target motor speed, acceleration and current loop feedforward compensation coefficients for the next control cycle.

2. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 1, characterized in that, In step S1, the method for obtaining the current vector sequence in the angle domain includes: The collected three-phase stator currents are transformed by Clark to obtain the current components in a two-phase stationary coordinate system. By combining the rotor position signal, the current components are transformed to the dq synchronous rotating coordinate system through Park transformation to obtain the quadrature axis current and the direct axis current. For the dq axis current signal, linear interpolation is performed based on the rotor position signal, and resampling is performed at preset fixed angle intervals to obtain the current vector sequence in the angle domain.

3. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 2, characterized in that, The linear interpolation based on the rotor position signal is achieved by using a cubic spline interpolation algorithm.

4. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 1, characterized in that, In step S1, the method for obtaining the composite feature matrix includes: The cross-axis current sequence in the angular domain is processed using short-time Fourier transform, and a time-frequency spectrum is generated. Based on the gear meshing frequency and bearing natural frequency of the seat drive chain, the time spectrum is divided into three frequency bands: the load change band, the gear meshing band, and the bearing friction band, which are respectively the frequency bands that increase sequentially. The Shannon time-frequency entropy of the energy distribution in each frequency band is calculated, and the fractal dimension of the signal in each frequency band is calculated by the box-counting dimension method. The six eigenvalues, including the Shannon time-frequency entropy and the fractal dimension of each frequency band, are combined into a composite eigenvector. When the seat motor is running continuously, the composite feature vectors are continuously generated, and these continuous composite feature vectors are arranged in chronological order to form a composite feature matrix.

5. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 4, characterized in that, The frequency range of the load variation band is 0-5Hz, the frequency range of the gear meshing band is 50-150Hz, and the frequency range of the bearing friction band is 200-500Hz.

6. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 1, characterized in that, In step S2, the method for calculating the posterior probability includes: The implicit state chain of passenger load includes three states: no load, standard load, and overload; the implicit state chain of mechanical resistance includes two states: normal resistance and increased resistance. Using a forward-backward algorithm, based on the observed composite feature matrix, the posterior probability of each possible state in the two parallel hidden state chains of passenger load and mechanical resistance is iteratively calculated at the current moment.

7. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 1, characterized in that, In step S3, the adjustment time is the total time it takes for the seat to move from the starting position to the target position. ; The driving torque fluctuation rate is the standard deviation of the quadrature axis current sequence during the adjustment process; The predicted temperature rise is based on the motor winding resistance. Heat capacity and the root mean square current during the adjustment process Calculated temperature rise Temperature rise The calculation formula is: .

8. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 6, characterized in that, In step S3, the current state distribution is the current joint probability distribution of the seat system obtained by multiplying the posterior probabilities of passenger load and mechanical resistance; the nominal distribution is the joint probability distribution of the seat system under ideal conditions; wherein, the probability corresponding to the joint state of standard load and normal resistance is 1, and the probability corresponding to all other joint states is 0; the divergence is the degree of deviation between the current joint probability distribution of the seat system and the joint probability distribution under ideal conditions.

9. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 1, characterized in that, In step S3, the method for defining the initial population search domain of the non-dominated sorting genetic algorithm using the posterior probability includes: If the posterior probability that the passenger load is in an overloaded state is greater than the first confidence threshold, then the upper limit of the initial search range for the target speed and acceleration of the motor is reduced, and the upper limit of the initial search range for the current loop feedforward compensation coefficient is increased. If the posterior probability of the mechanical resistance being in an increasing resistance state is greater than the second confidence threshold, then the initial search lower bound of the current loop feedforward compensation coefficient is increased.

10. The method for dynamically optimizing seat adjustment parameters in a commercial vehicle according to claim 9, characterized in that, The first confidence threshold is set to 0.8, and the second confidence threshold is set to 0.

7. If the posterior probability that the passenger load is in an overloaded state is greater than 0.8, then the lower limit of the initial search range of the motor target speed will be reduced by 20%, the upper limit of the initial search range of acceleration will be reduced by 15%, and the upper limit of the initial search range of the current loop feedforward compensation coefficient will be increased by 30%. If the posterior probability of the mechanical resistance being in an increasing resistance state is greater than 0.7, then the lower limit of the initial search range of the current loop feedforward compensation coefficient is increased by 25%.

Images