Research And Implementation Of Printing Equipment Intelligent Diagnosis System Based On Deep Learning

At present,the annual output value of my country's printing,packaging and publishing industry is about 3 trillion yuan.It is already a big printing country,a big packaging country and a big publishing country.The failure of printing equipment has greatly affected the production efficiency.With the continuous transformation of traditional manufacturing enterprises to the direction of high quality and high efficiency,along with the progress of science and technology and the development of industrial big data,the integration of artificial intelligence technology and advanced manufacturing technology has created a new way of intelligent manufacturing,which has become a new round of industrial revolution.The backbone,making it possible to intelligently diagnose printing equipment.Based on this background,this thesis studies a printing equipment diagnosis system based on deep learning,proposes an improved algorithm for bearing health status assessment and remaining service life prediction,realizes real-time monitoring of printing equipment operating parameters and proposes an information-based equipment point It can improve the production efficiency,strengthen the safety of equipment,and reduce labor costs.The improved algorithm analyzes the vibration signal data generated during the operation.This thesis takes the rolling bearing commonly used in printing equipment as the research object to describe the experiment.The main research contents are summarized into the following three parts.(1)Deep Residual Shrinkage Network(DRSN)are used for bearing health status assessment.This model has anti-noise ability and can learn features from strong noise signals,and can obtain status assessment results that are more suitable for the actual situation.The soft thresholding is added to the neural network as a nonlinear layer,and the attention mechanism is introduced to automatically set the threshold from the network to achieve the purpose of removing unimportant features.This method gets rid of the dependence of traditional feature extraction methods on expert experience,the data does not need to be professionally processed,and the neural network can automatically learn features from the original data.Finally,it is verified on the XJTUSY Bearing Datasets.The experimental results show that the bearing health state assessment method based on the deep residual shrinkage network can better reflect the change process of the entire life cycle of the bearing,which is helpful to improve the follow-up the accuracy of the forecast work.(2)A method for predicting remaining useful life based on Long-Short Term Memory(LSTM)network is improved.On the basis of evaluating the bearing health status and establishing the bearing health factor,the training samples of the neural network are constructed by the time window sliding method,and combined with the advantages of the LSTM network in processing time series data,the purpose of predicting the bearing health status in the next stage is achieved.In this thesis,a singlelayer LSTM structure,a multi-layer LSTM structure and a bidirectional LSTM structure network are built,and a comparative experiment is carried out on the influence of the network structure on the prediction of the remaining service life of the bearing.The experimental results on the public data set show that the prediction of the bidirectional LSTM network The effect is better than the single-layer LSTM network and the multilayer LSTM network,and the multi-layer network structure achieves higher prediction accuracy than the shallow network structure.(3)The intelligent diagnosis system of printing equipment with B/S structure was built,and the hardware and software parts were overall planned.In the hardware part,an acceleration sensor and a temperature sensor are installed on the printing equipment to collect industrial data,and the data is transmitted through the Modbus communication protocol to build a real-time operating parameter database of the printing equipment.The software part builds and deploys a deep learning model through the Keras framework,uses the front-end and back-end separation technology to build system services,calls the trained deep learning model,and realizes functional modules such as management,spot inspection scheme configuration,equipment operation monitoring,and fault warning.