Search

CN-122008759-A - Full-active suspension system control method based on rule learning

CN122008759ACN 122008759 ACN122008759 ACN 122008759ACN-122008759-A

Abstract

The invention discloses a control method of a full-active suspension system based on rule learning, which comprises the steps of 1 collecting dynamic response parameters of vehicle running under different working conditions through a vehicle dynamics model to form a vehicle state data set, 2 carrying out cluster analysis and labeling on the vehicle state data set by utilizing a k-means algorithm to form a data set containing cluster labels and modes, 3 introducing a RIPPER algorithm in rule learning to carry out rule extraction and pruning optimization to output a final rule set, and 4 designing a suspension system control strategy based on rule learning aiming at a semi-active suspension mode and a full-active suspension mode. The method combines rule learning with mode switching control of the full-active suspension system, formulates interpretable rules and switching thresholds, has the advantages of rule interpretability and small calculated amount, and is beneficial to realizing mode switching and optimal control of the suspension system so as to improve the comfort and control stability of the vehicle under complex working conditions.

Inventors

  • LI ZHIYING
  • BAI XIANXU
  • LI JIE
  • ZHU ANDING

Assignees

  • 合肥工业大学

Dates

Publication Date
20260512
Application Date
20260130

Claims (5)

  1. 1. The full-active suspension system control method based on rule learning is characterized by comprising the following steps of: Step 1, inputting road surface excitation into a vehicle dynamics model, obtaining relevant state parameters of the vehicle dynamics model, and forming a vehicle state data set Wherein, the method comprises the steps of, Represent the first A sequence of individual vehicle state response parameters, an , Representation of Middle (f) The state of the strip vehicle is that, Representation of In the total number of vehicle states, Representing a total number of vehicle state response parameter sequences; Defining the mode of an all-active suspension system as If the mode of the all-active suspension system is a semi-active suspension mode, the method causes If the mode of the all-active suspension system is the all-active suspension mode, then ; Order the In (a) The first of the individual vehicle state response parameter sequences The mode of the optimal all-active suspension system corresponding to the state of the vehicle is recorded as Thereby obtaining the mode sequence of the optimal all-active suspension system corresponding to each vehicle state ; Step 2, adopting k-means algorithm to make the first step Individual vehicle state response parameter sequences Clustering to obtain Is a result of clustering of (a) Wherein, the method comprises the steps of, Represent the first The individual vehicle state response parameter sequence clusters, Representing the clustered cluster number, and The serial number j and the cluster serial number u of the corresponding vehicle state response parameter sequence are used as cluster labels and marked as Thereby obtaining a cluster label set The maximum value and the minimum value of each vehicle state response parameter sequence are taken as Is a cluster range of (a); step 3, the mode sequence Matching with cluster label set to form data set containing cluster label and mode ; Step 4, adopting rule learning RIPPER algorithm pair Extracting potential rules in the rule set to form a final rule set Wherein, the method comprises the steps of, Represent the first The rule of the strip optimization is set, Representing default rules, N representing the total number of optimization rules; Step 5, acquiring the full-active suspension system before the time t History pattern of each And calculate the suspension mode variation between two adjacent history modes if The change amount of each suspension mode is kept unchanged, and then a switching signal at the moment t is output Otherwise, outputting a switching signal at the time t ; Step 6, obtaining the t time The first vehicle state response parameter sequence Strip vehicle status And is connected with Mapping the cluster range and then combining with Matching to obtain the suspension mode at the current t moment And suspension mode after time t ; Step 7, combining And And (3) with Final demand mode of output all-active suspension system To achieve control of the fully active suspension system.
  2. 2. The method for controlling an all-active suspension system based on rule learning according to claim 1, wherein the step 4 is performed as follows: step 4.1, according to the set extraction rule' ", From Extracting different rules to form a preliminary rule set , wherein, Represent the first The rule is a preliminary rule of the strip, Representing the total number of preliminary rules; step 4.2-will be As a positive example, will Is taken as a counterexample, thereby calculating the performance metric index of the preliminary rule for the preliminary rule set Pruning and optimizing to form an optimized rule set , ; Step 4.3, setting a default rule " ", For processing Cluster labels not covered in (a) to form a final rule set Wherein the conditions are Representation of Cluster labels not covered in (a).
  3. 3. The method for controlling an all-active suspension system based on rule learning according to claim 1, wherein step 7 comprises: if it is And (3) with Is the same, not to Switching, directly to As a final demand pattern Output to suspension controller, otherwise, when When it will As a final demand pattern Output to suspension controller when When it will As a final demand pattern Output to the suspension controller.
  4. 4. An electronic device comprising a memory and a processor, wherein the memory is for storing a program that supports the processor to perform the method of any of claims 1-3, the processor being configured to execute the program stored in the memory.
  5. 5. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when run by a processor performs the steps of the method according to any of claims 1-3.

Description

Full-active suspension system control method based on rule learning Technical Field The invention relates to the field of rule learning and suspension multi-mode switching, in particular to a full-active suspension system control method based on rule learning. Background In the running process of a vehicle, the vehicle often faces external complex and changeable working conditions, is excited by vibration from a road surface, has the characteristics of multiple frequency bands and time variation, and has extremely high requirements on the smoothness and the operation stability of a suspension system. Although some advances have been made in suspension control technology research, for example, optimal control algorithms such as model predictive control (model predictive control, MPC) exhibit some effects in suspension multivariable optimization, there are still limitations in the area of multi-mode switching control. The development of suspension systems has significantly improved the handling capacity of vehicles on various terrains, and good suspension systems can improve the smoothness or steering stability of vehicles, and common vibration reduction systems are divided into passive suspensions, semi-active suspensions and all-active suspensions. The passive suspension has simple structure and low cost, but the parameters of the passive suspension cannot be adjusted, and the vibration reduction effect is limited. The full-active suspension can generate acting force with adjustable size and direction, and has good vibration damping performance, but the full-active suspension is difficult to popularize due to the factors of high energy consumption, complex structure, overhigh cost and the like. Semi-active suspension performance is between passive and full-active suspension performance, but its actuator has limited mechanical range, thus limiting its optimal control effect in the full frequency range. Due to the complexity and variability of the running conditions of the vehicle, the difference of the requirements on the suspension performance under different conditions is remarkable, and the existing control scheme generally adopts a single control strategy or simple multi-mode switching logic to cope with the vibration problem, so that the dynamic change of multiple conditions cannot be adapted. Meanwhile, the existing suspension multi-mode switching logic only sets switching conditions according to single state parameters, such as a vehicle body acceleration threshold value, cannot mine the internal correlation between the state and the working condition of the multi-dimensional vehicle, and cannot form a clear and interpretable mode switching boundary. The switching decision is based on fuzzy and low logic transparency, so that the requirements of suspension performance under different working conditions are difficult to precisely match, the deviation of the mode switching time is easily caused by unclear boundaries, the comfort of passengers is influenced, and potential risks are caused to the stability of a vehicle system and the safety of passengers. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a full-active suspension system control method based on rule learning, so that the mode switching and optimal control of a suspension system can be realized, the self-adaptability, the interpretability and the instantaneity of the suspension system can be enhanced, and the comfort and the control stability of the suspension under complex working conditions can be optimized. The invention adopts the following technical scheme for solving the technical problems: The invention relates to a full-active suspension system control method based on rule learning, which is characterized by comprising the following steps: Step 1, inputting road surface excitation into a vehicle dynamics model, obtaining relevant state parameters of the vehicle dynamics model, and forming a vehicle state data set Wherein, the method comprises the steps of,Represent the firstA sequence of individual vehicle state response parameters, an,Representation ofMiddle (f)The state of the strip vehicle is that,Representation ofIn the total number of vehicle states,Representing a total number of vehicle state response parameter sequences; Defining the mode of an all-active suspension system as If the mode of the all-active suspension system is a semi-active suspension mode, the method causesIf the mode of the all-active suspension system is the all-active suspension mode, then; Order theIn (a)The first of the individual vehicle state response parameter sequencesThe mode of the optimal all-active suspension system corresponding to the state of the vehicle is recorded asThereby obtaining the mode sequence of the optimal all-active suspension system corresponding to each vehicle state; Step 2, adopting k-means algorithm to make the first stepIndividual vehicle state response parameter sequencesClusterin