On-line Diagnosis Method of In-wheel Motor Mechanical Fault Based on Dynamic Bayesian Network

Abstract

The invention discloses an on-line diagnosis method for mechanical faults of a hub motor based on dynamic Bayesian network. The method comprises the following steps: respectively calculating time-domain and frequency-domain high-sensitivity characteristic parameters of each segment of acceleration signals, constructing a Bayesian network structure of a velocity slice vk corresponding to the current time segment k, obtaining the conditional probability distribution of each of network nodes, determining state transition probability distributions of a secondary velocity slice vk-1-vk between thevelocity slice vk-1 corresponding to the previous time segment k-1 and the velocity slice vk corresponding to the current time segment k, establishing a dynamic Bayesian network model, establishing afault diagnosis model group based on a plurality of state transition probability distributions between the velocity slice vk-1 and the velocity slice vk, collecting the operation information of the hub motor online, selecting a corresponding diagnosis model of the secondary velocity slice vk-1-vk from the fault diagnosis model group, calculating the posterior probability distribution, and determining whether the hub motor is normal or faulty, thereby improving the accuracy and timeliness of identifying mechanical faults of the hub motor.

Description

technical field [0001] The invention relates to the field of state monitoring and intelligent diagnosis of mechanical faults of automobile hub motors, in particular to an online diagnosis method for mechanical faults of hub motors based on a dynamic Bayesian network. Background technique [0002] Electric vehicles driven by in-wheel motors have outstanding advantages such as simple and compact structure and high transmission efficiency. The in-wheel motor is the core of the electric vehicle drive system. Since the in-wheel motor is installed in the narrow hub space, factors such as magnetic field saturation, torque fluctuations, and load mutations have a significant impact on its performance, and the changing driving conditions and complex road conditions It is very easy to induce mechanical failure of the wheel hub motor, which will cause increased vibration, decreased efficiency, and increased temperature rise. Continuous long-term operation will also lead to decreased per...

AI Technical Summary

Problems solved by technology

[0004] (1) In 2013, Li K et al. mentioned a discrimination index method (Distinguish index, DI), which consists of DI value and two-state discrimination rate (Discrimination Rate, DR), but this method is only fast and effective for a single high-sensitivity feature parameter in two states, and cannot select multiple high-sensitivity feature parameters at the same time

[0005] (2) In 2012, Li K et al published a comprehensive discrimination index method ( Synthetic detection index, SDI), is used to extract multiple sensitive feature parameters, which is composed of DI values ​​​​under multiple states, but this method ignores that mechanical equipment usually has multiple working conditions, and it cannot select multiple working conditions. Highly Sensitive Characteristic Parameters

[0007] (1) In 2011, in the paper "Fault Diagnosis of Motor Bearing Based on the Bayesian Network" in the journal "Procedia Engineering" (Volume 16, No. 16), Li Z et al. published a method of applying Bayesian network to motor bearing faults. The diagnosis method is based on the vibration signal and the Bayesian diagnosis model, but this method does not take into account the influence of the previous time period on the current time period, and cannot dynamically adjust the diagnosis results

[0008] (2) In 2016, Zhao Yuenan et al. published a paper using cuckoo algorithm in the paper "Application of Bayesian Network Using Cuckoo Algorithm in Asynchronous Machine Fault Diagnosis" in the magazine "Mechanical and Electrical Engineering" (Volume 33, No. 2). The Bayesian network of the algorithm is applied to the fault diagnosis of asynchronous motors. It diagnoses based on the fault current signal of the asynchronous motor and the Bayesian diagnosis model. However, due to the complex operating environment of the hub motor, this method has no The identification is relatively low, and this method is a fault diagnosis based on offline data, which cannot achieve fast and effective online identification and diagnosis, and cannot meet the operation safety requirements of in-wheel motors

Images