Research On Equipment Reliability Evaluation And Data Security For Industrial Internet

Industrial Internet is the core of the fourth industrial revolution and to promote manufacturing industry to achieve high level intelligent transformation.The development of Industrial Internet has risen as a national strategy in China.As the main production body of intelligent manufacturing,the reliability and security of industrial equipment and the privacy protection of industrial big data are important to ensure the efficient operation of industrial systems.In this paper,the research on the remaining useful life of turbine engines is done to explore the impact of data-driven approach on reliability assessment of industrial equipment in the era of industrial Internet,as well as the data privacy and security protection problems faced by the application of industrial big data.Based on classical three-parameter Weibull distribution and deep learning neural network LSTM,the remaining useful life prediction models of turbine engines are established.By comparing the results,we found that the traditional reliability evaluation model(Weibull)is lacking of flexibility and unable to deal with mass real-time data.Industrial equipment using this method is difficult to guarantee the safe operation even with the real-time condition monitoring service.The deep learning method performs well in degradation characterizing.Without knowledge of degradation mechanism and experts experience,it can still maintain good processing ability when facing the increase of system complexity.However,data-driven methods may cause data leakage when processing the industrial data,so the protection of industrial big data security is essential to promote the construction and development of Industrial Internet security.In this context,the federated learning method is introduced in this paper.By cooperating the data,the precision of remaining useful life prediction is dramatically improved where the recall rate of each participant is improved by 40% and 61% respectively.At the same time,a data leak model is established,and the gain of model accuracy improvement is quantified with maintenance cost.The applicability of federated learning compared with local learning and centralized learning is discussed in terms of data security protection and model accuracy improvement.The results of this paper can provide reference for equipment reliability assessment and security protection of industrial big data in the era of Industrial Internet,and provide an effective scheme for accelerating the development of Industrial Internet and realizing intelligent,digital,networked transformation and high-quality development of manufacturing industry.