Equipment Health Condition Monitoring Technology Using Entropy

Unexpected problems and failures usually occur during the operation of industrial Machinery, and the timelydiscovery caused great loss of human and financial resources. Therefore, it is significant that timely monitoring ofthe health status of the device, making it operate safely. In this paper, studies the structural health monitoringtechnology based on the amount various features of entropy.The paper has studied the method of the signal pretreatment, such as the time-domain characteristic,frequency-domain characteristic and time-frequency domain characteristic. Experimental results show that, Whenrolling bearing failure, time domain, frequency domain characteristic will change. There are differences betweenthe different characteristic, When occurred different types of damage or different degrees of damage, the time andfrequency domain indexes will have significant differences. Kurtosis coefficient has the biggest differences intime-domain characteristic, gravity frequency has the biggest differences in frequency-domain characteristic.Inaddition, the different states vibration signals after wavelet packet decomposition, its energy distribution is verydifferent too. Therefore, extract the. vibration signal of the time-domain characteristic, frequency-domaincharacteristic and time-frequency domain characteristic can reduce the dimension of the vibration signal,effectively describe the different types of fault condition.The health status monitoring methods based on frequency-domain characteristic of the signal entropy hasbeen studied in paper. On the basis of the extracted signal characteristics,discussed the effectiveness of Shannonentropy, approximate entropy, sample entropy and Permutation entropy monitoring equipment and structuralhealth state. Proved by experiments that entropy algorithm can be used to monitor the health status of differentstructural health, and the frequency-domain characteristic can effectively describe the condition and monitoringequipment and structures.The structure damage is a process of increasing, In order to monitor the entire damage process moreeffectively and predict the damage, and reduce the loss caused by damage to a minimum. The paper studied theself-regression model to monitor equipment and structural health. The equipment and engineering structure dataexperiment indicate that, self-regression model can better fit the changing trends and forecasts of entropy.In order to overcome the problem that self-regression model has a large error in non-linear data prediction,studied the approach of health status monitoring and prediction based on neural network. The experiment showthat: the neural network method can accurately predict the state and health trends of machinery and engineering structures.