Application of Big Data Analysis for Fault Diagnostics in Maintenance

Since recent advancements in bigdata seek to extract data, information, and knowledge from huge records, they have attracted a lot of attention. The development of Information & Communication Technologies (ICT) and sensor technology in the industrial sector has allowed for the capture and storage of enormous volumes of streaming, and nonlinear records to keep decision making. Identification of issues to records is an essential component of e-Maintenance systems since it can aid in support result. Advanced system fault detection can help to reduce the risk of unexpected breakdowns and assure the dependability and safety of industrial systems. Far above the ground dimensionality may result in the infamous “curse of dimensionality” and worsen the precision of defect finding systems, according to the standpoint of data modeling. Algorithms must respond to the arrival of fresh samples in close to real time in order to handle fast-moving data streams. High nonlinearity necessitates that fault detection methods have adequate expressive capacity and steer clear of over fitting or under fitting issues. In this study, an ABSAD (Angle based Subspace-Anomaly Identification) method to detect in data of high-dimension is developed. Analytical analyses and numerical examples illustrate how effective the approach is. The come up to expanded to an available type constructed on the downhill pane strategy in order to find errors in high-dimensional data streams. The online extension is able to adapt to the time-varying behavior, according to experiments on synthetic datasets.