Research Of Tendency Prediction In Equipment Failure Based On Cloud Computing

With the development of modern industry, big data has become the important strategic resources of enterprises. How to safely save and share big data resources and excavate its potential value have been an imperative problem for enterprises. Cloud computing has put forward a new service mode about IaaS, PaaS, SaaS, which can be adapted for the demand of enterprises different stages, and it also provides a new model for the development of modern industry.This paper combines the cloud computing with equipment maintenance system, proposes the framework of equipment maintenance system cloud computing based on Hadoop, EM-HDFS, EM-MapReduce. The paper also respectively discusses equipment maintenance resource layer, equipment maintenance services layer and equipment maintenance application layer in detail. This paper focuses on the fault trend prediction of equipment maintenance services layer. This paper uses SVR to the failure prediction and analyzes the different effects of the parameters on the performance of support vector regression machine (SVR). This paper uses uses particle swarm optimization algorithm for parameter optimization of SVR. Using a standard set of UCI database data, this paper conducts the optimization experiment.In fact, the data size gradually increases towards large scale, the time required for traditional SVR in single cases increases sharply. In the light of these existing problems, this paper proposes a distributed SVR algorithm based on Hadoop. Experiments show that, in the case of almost same prediction accuracy by distributed SVR based on Hadoop and single SVR, the former greatly saves the computation time. And, at the same time, stay the sample data unchanged, this paper analyses the influence on time consumption by increasing the number of Map tasks. It turns out that increasing the number of Map tasks within a certain range can Reduce the time consumption.This paper builds the equipment fault trend prediction model based on Hadoop and distributed SVR, and uses the data collected from the mine as experiments proof for the model validation. The experiment proved that the application of distributed SVR based on Hadoop in the failure prediction not only has the advantages of saving time, high prediction accuracy and reliability, it’s prediction performance can also meet our requirements.