| CPC B60G 17/0185 (2013.01) [B60G 2202/413 (2013.01); B60G 2202/42 (2013.01); B60G 2600/08 (2013.01)] | 8 Claims |
|
1. A method for fault diagnosis of electro-hydraulic servo active suspensions, wherein an experimental platform for the electro-hydraulic servo active suspensions is adopted, and the experimental platform comprises a gantry frame mainly comprising a bottom plate and vertical columns, input signal actuators, a wheel suspension apparatus, a sprung mass block and an inertial sensor are sequentially arranged on the bottom plate of the gantry frame from bottom to top, an electro-hydraulic servo actuator is arranged on the wheel suspension apparatus, an upper part of the gantry frame is provided with counterweight blocks for increasing a weight of the sprung mass block, and a computer is arranged at one side of the gantry frame;
the wheel suspension apparatus comprises wheels, shock absorbers, hydraulic cylinders, and cylinder body vibration sensors, the shock absorbers are arranged at upper parts of the wheels, the hydraulic cylinders are arranged behind the shock absorbers, the cylinder body vibration sensors are arranged on the hydraulic cylinders, and the sprung mass block is arranged above the shock absorbers and the hydraulic cylinders;
the electro-hydraulic servo actuator comprises an oil tank, a hydraulic pump, valve body vibration sensors, servo valves, and a pump body vibration sensor, the hydraulic pump is in communication with a lower part of the oil tank, the oil tank and the hydraulic cylinders at two sides form an oil inlet circuit and an oil return circuit, respectively, the hydraulic pump and the servo valves are sequentially arranged on the oil inlet circuit, the pump body vibration sensor is arranged on the hydraulic pump, and the valve body vibration sensors are arranged on the servo valves; and
the method for fault diagnosis comprises the following steps:
S1: inputting an actuating signal to the wheel suspension apparatus by the input signal actuators;
S2: adjusting and controlling the electro-hydraulic servo actuator according to a control solution;
S3: monitoring a longitudinal displacement and a lateral inclination angle of the sprung mass block by the inertial sensor, recording longitudinal displacement and lateral inclination angle data, and forming a curve of the longitudinal displacement and the lateral inclination angle changing with time, wherein a time interval of data recording is T;
S4: detecting vibration signals of the servo valves, the hydraulic cylinders and the hydraulic pump in an actuating state respectively by the valve body vibration sensors, the cylinder body vibration sensors and the pump body vibration sensor, and respectively recording the vibration signals according to different hydraulic elements, wherein a time interval of vibration signal recording is T, and recording beginning and ending time is consistent with that of the longitudinal displacement and the lateral inclination angle;
S5: performing noise reduction processing on acquired vibration signals by combining with acquired longitudinal displacement and lateral inclination angle data as well as positional relationships between vibration sensors of the different hydraulic elements and the inertial sensor;
S6: analyzing vibration signals after noise reduction respectively in a time domain and a frequency domain, and then extracting feature data of the vibration signals;
S7: respectively constructing fault feature data sets according to the different hydraulic elements and fault forms corresponding thereto;
S8: judging whether a fault feature data volume of the different hydraulic elements meets actual fault diagnosis requirements; if not, changing an input actuating signal form or increasing the counterweight blocks on the sprung mass block or repeating an experiment multiple times to acquire fault feature data under a same working condition, jumping to S1, and continuing to collect fault feature data of the different hydraulic elements; and if yes, proceeding to a next step;
S9: judging whether vibration signals under different fault forms in the different hydraulic elements are all recorded, and constructing fault feature data sets; if a hydraulic element having a set fault form is not recorded, correspondingly replacing a hydraulic element having a same model and having the set fault form in the experimental platform, and jumping to S1, thereby constructing a feature data set of the set fault form; otherwise, proceeding to a next step; and
S10: respectively importing the fault feature data sets of the different hydraulic elements into a deep learning model for training to obtain fault diagnosis models of the different hydraulic elements,
wherein in actual fault diagnosis, within a time interval T, a longitudinal displacement and a lateral inclination angle of a vehicle body and an actual vibration signal of a hydraulic element are monitored, equivalent processing in step S5 and step S6 is performed on the actual vibration signal, and a processed actual vibration signal is input into a fault diagnosis model of a corresponding hydraulic element to judge whether an actual hydraulic element has a fault, thereby completing fault diagnosis of active suspensions; and
a method for noise reduction processing of vibration signals in step S5 is as follows:
assuming that acquired longitudinal displacement data is represented by a function D(t), in which an initial position is zero and upward is positive, and lateral inclination angle data is represented by a function φ(t), in which an initial position is zero and clockwise is positive; and assuming that a straight line distance between a vibration sensor of a certain hydraulic element and the inertial sensor is 1 and a horizontal included angle is θ, and a vibration signal monitored by the vibration sensor is L(t), a vibration signal generated by the vibration sensor under an influence of the sprung mass block is capable of being approximately D(t)+1*sin [φ(t)−θ]; and
a vibration signal after noise reduction, M(t)=L(t)−D(t)−1*sin [φ(t)−θ]−ψ[L(t), D(t), φ(t)], is capable of being obtained, where ψ[L(t), D(t), φ(t)] is a phase error correction term of L(t), D(t) and φ(t).
|