Research On Equipment Behavior Recongition Based On Hidden Markov Chain In Industrial Big Data Environment

As the scale of the enterprise grows and the degree of automation improves,the management scope of managers becomes larger and larger,and decision-making becomes more and more complicated.This requires managers to quickly grasp the equipment status and operating behaviors on the shop floor.The traditional method for managers to visit the site to understand the situation is inefficient,at the same time the behavior and status of equipment are not traceable.With the development of the Industrial Internet,equipment operation data can be sensed in real time,which provides a basis for managers to automatically obtain equipment status and operation behavior.The purpose of this paper is to analyze and process the equipment operation data based on real-time perception,then identify the equipment status and operation behavior,which hidden behind the data.This is for automatically feed back to the manager as a decision reference to promote transparent management of production.The main contents of this paper as follow:(1)After perform outlier processing on the obtained equipment operation data,use basic statistical index analysis,reconstruction and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise analysis,and the maximum likelihood probability of hidden Markov chain for feature extraction.Select the optimal feature from the extracted features,then use two methods,which separately use each optimal feature and combine each optimal feature,to evaluate the effect of the hidden Markov chain equipment behavior recognition model.Use the results of the optimal model as the final equipment behavior.Compared with the running time,the accuracy and recall of equipment behavior recognition of K-means,agglomerative clustering,DBSCAN and GMM,which are four commonly used unsupervised learning models,in order to highlight the superiority of hidden Markov chain device behavior recognition model.(2)The modeling process method for equipment behavior recognition and the hidden Markov chain itself are processed by Map Reduce,and the effect is evaluated by two performance indicators,namely acceleration ratio and scalability.Experimental results show that modeling process method and hidden Markov chain of equipment behavior recognition after Map Reduce processing can effectively improve the continuous learning efficiency and reduce the corresponding processing time of massive historical equipment operation data.(3)On the basis of the existing energy management system,two submodules of the hidden Markov chain equipment behavior recognition model and the hidden Markov chain equipment behavior recognition Map Reduce parallel model are added,and the corresponding equipment behavior recognition interface is completed.