Recognition Method Of Working State Of Washing Machine Based On Multi-parameter Feature Fusion

The national standard GB/T 4288-2018 has put forward strict requirements on the dehydration speed,vibration performance and noise of household and similar electric washing machines.In order to solve the problem of dependence and complementarity between different parameters in the dewatering section,this paper proposes to extract 3D vibration signals from the outer box of the dewatering section.The three parameters of vibration,current and sound in the process of dehydration were detected at the same time.The feature extraction and fusion of multiple parameters in different working modes were carried out to achieve the purpose of working mode recognition in the dehydration section.The main research work in this article is as follows:(1)Rotation speed extraction from 3D vibration signals in the dehydration stage based on improved segmental AMDF.The desired speed curve of the drum is extracted from the STFT-based maximum harmonic value of the motor current,and the speed curve is segmented by the adaptive first-order-difference-threshold method.Design an adaptive differential amplifier circuit to improve the signal to noise ratio of the rising vibration signal.The segmented Average Magnitude Difference Function(AMDF)is used to extract the rotational speed of the vibration signal of shell and outer tub.Taking six target speeds of a certain type of drum washing machine in the dehydration stage as the experimental object,set the sampling rates as 10 k Hz,1k Hz and 500 Hz respectively to verify the method in this paper.The experimental results show that the rotation speed extraction method proposed in this paper has low sampling rate versatility and 3D(3-axis direction)robustness compared to earlier single sampling rate(10k Hz)and single direction(signal strongest direction)extraction methods.(2)Classification of dehydration stage work modes based on vibration signal rotation speed curves.Rotation speed extraction based on segmentation of desired drum speed curve and AMDF.Six target rotational speed operating modes in the dehydration stage are classified using DTW,KNN,SVM,and LSTM classification methods.The experimental results show that the classification and recognition rate of vibration signal rotational speed curve based on KNN is the relatively optimal.(3)Classification of dehydration stage work modes based on multi parameter feature extraction and fusion.The vibration,sound,and current parameters of the dehydration stage are extracted separately.To solve the problem of large classification model time overhead caused by excessive dimension of combined features,KPCA is used to reduce the dimension of combined features.Using KNN,SVM,and LSTM classification methods,classification experiments are conducted on fused features without dimensionality reduction and after dimensionality reduction by KPCA.The experimental results show that when the current signal extracts wavelet packet features,the sound signal extracts MFCC features,and the vibration signal extracts EMD energy entropy,the KNN classification method is used to classify the above non dimensioned feature combinations,and the recognition rate is relatively highest;However,considering the time cost comprehensively,the LSTM classification method is relatively effective.Figure [58] Table [12] Reference [61]...