Research On Condition Monitoring For The Brake Pads Of Civil Aircraft By Data-driven Approaches

As the core mechanism of civil aircraft braking system,brake pads are multi-input and multi-output control system that highly coupled with mechanical-electro-hydraulic.They have serious nonlinear characteristics.Because of the complex structure and extremely harsh working environment,brake pads have a high failure rate.The accurate and effective monitoring of the condition of the brake pads can help to determine the performance of brake pads,and to find potential failures in time to avoid serious accidents.Therefore,it is of great theoretical and practical significance to study the condition monitoring technology of civil aircraft brake pads.This thesis conducts in-depth research on the brake pads of civil aircraft.A new fusion algorithm is formed by improving and integrating support vector regression(SVR)algorithm and particle filter(PF)algorithm from the perspective of data-driven method.The friction torque and wheel speed are estimated by the fusion algorithm to monitoring the condition of the brake pads.The specific work is as follows:First,the kernel function of the SVR algorithm is analyzed.It is found that the tradition kernel function was only considered the relationship between sample data variables,and was ignored the distribution information of the sample itself.Based on this,a newly probability kernel function containing sample data distribution information is proposed from the perspective of the construction principle of the kernel function to improve the output accuracy of SVR algoritnm,and pave the way for the subsequent improvement of the resampling technology of PF algorithm.Secondly,in view of the disadvantage of particle's degeneracy phenomenon in the standard PF algorithm,a method for selecting the importance probability density function is designed based on Hellinger distance.Consider to maximizing the similarity between the newly designed importance probability density function and the theoretically optimal importance probability density function,that is,to minimizing the Heringer distance between the two density functions.The improved importance probability density function is calculated to alleviate the particle's degeneracy phenomenon of the standard PFalgorithm.Then,aming at the shortcoming of particle's impoverishment of the PF algorithm,a resampling technology incorporating the SVR algorithm is designed.The proposed probability kernel function is used to construct the probability density function of particles when the PF algorithm is resampled,and this probability density is obtained through the SVR model.Finaly,the probability density function is used to calculate the particles and their corresponding weights during resampling.Therefore,a new resampling technology is proposed,which can alleviate the particle's degeneracy phenomenon and ensure the diversity of particles,and finally forms a new fusion algorithm.The state variables of the brake pads are estimated by fusion algorithm,so as to achieve the purpose of condition monitoring of the brake pads.Finally,a simulation environment is constructed,and the data of friction torque of the brake pad and the wheel speed are obtained to evaluate and verify the overall performance of the SVR-PF algorithm.