Research On WPT-DDT-SVM And Application To The Fault Diagnosis Of The Hydraulic Pump Of Concrete Pump Trucks

Hydraulic pump has been widely used in the modern construction machinery and many other mechanical equipment, Statistics show that about30to40%of pump truck’s faults are concerned with the construction machinery. It is important to monitor hydraulic pump running state and predict early failure for reducing the faultrate and downtime of the hydraulic engineering machinery.Taken this as the background, a hydraulic pump fault diagnosis method was put forward systematically, and the key was to improve the performance of fault classification.In order to overcome the frequency mixing of wavelet packet transformation, band adjusted wavelet packet reconstruction algorithm was used for feature extraction.Experiments demonstrate this method can eliminate the mixing band effectively.Aaiming at the irrelevant and redundant information in the feature space, the distance discriminant technique was used to select features. Experiments demonstrate that DDT can can reduce computational complexity and improve the classification accuracy extremely.According to the poor performance of the traditional method in large dataset and strong noise environment, the various kinds of fault class and atrocious work condition of hydraulic pump, a novel state recognition method called fault-tolerant adaptive support vector machine (FTASVM) was proposed in this paper. It achieved a fast classification by:(1)importing fault-tolerant;(2) selecting the binary SVMs which can divide one class from all other classes;(3)selecting the binary SVMs with the fewest average number of support vectors (SVs);(4)To improve the adaptability of multi-fault diagnosis incremental learning algorithm was imported to train the model. Experiments demonstrate FTASVM can speed up the test phase remarkably and remain the high accuracy of classification.