Research On Identifying Lathes' State And Evaluating Workers Effectiveness Based On Vibration Analysis

The vibration signals of the device covers a lot of information, it is not only used of doing fault diagnosis but also used state recognition. Mechanical equipment vibrates differently with different working condition, the vibration signal is different also, using corresponding signal analysis method, distinguishing between different working states. By using three different methods to identify the states of a lathe, statistics the time of every state, doing some evaluation for workers and equipment with the time.(1) Based on BP neural network for data classificationLooking for the working states of he time-domain and frequency-domain features as network input, depending on the type and number of input and outputs'characteristics designing the structure of BP neural network.Using the frequency domain characteristics,frequency domain and the mixed together three kinds of data as the network input to classify data and statistics time of each working state. The results show that mixed features as input has highest success rate.(2) Based on ZCPA for data classificationThe improved ZCPA model used in extracting characters of mechanical vibration signal. Describes the basic principle of the model and based on the characteristics of vibration signals made the corresponding adjustment, adopt S function to simulate the human cochlea hair cells, use the Gammatone filter group to make model of cochlear basement membrane,the frequency of cases according to the frequency range of geometric incremental approach in the frequency distribution on the axis. Results show that using ZCPA model to extract state feature of five kinds, can work accurately distinguish them between different states. By calculating the Pearson correlation coefficient, can statistics out of the time of each working state.(3) Based on fuzzy theory for data classificationSignal amplitude mutation changes when working state changes. Through doing the half-wave rectifier and enveloping signals three times, establish corresponding fuzzy comprehensive threshold. Doing like that can accurately recognize the shocking signal, analysis of working state before and after the shocking point, find out the status changing point. The results show that using this method can accurately identify the different working state, extract the shocking signal snd statistics time of each working state.The time statisticed and the shock signal extracted can be used to provide the reference calculat equipment investment payback period and equipment maintenance period, do assessment of works' operation level and emotion, etc.