Using IaaS to deploy novel ANN for asset management of rotating equipment on sites across the globe

Journal of Computer Science & Systems Biology

ISSN: 0974-7230

Open Access

Using IaaS to deploy novel ANN for asset management of rotating equipment on sites across the globe

Joint Event on 5th World Machine Learning and Deep Learning Congress and World Congress on Computer Science, Machine Learning and Big Data

August 30-31, 2018 Dubai, UAE

Saeed Nusri

ATCO Group, Canada

Posters & Accepted Abstracts: J Comput Sci Syst Biol

Abstract :

With IaaS platforms like AWS readily available, it has become increasingly easy to deploy IoT platforms with prognostic capabilities. This availability has allowed a myriad of startups to easily enter and compete in the tech space with cutting edge and scalable machine learning technologies for a variety of applications. Some of these applications include asset management and condition monitoring of industrial equipment, which is a multi-billion dollar industry. This presentation focuses on showing how IaaS can be leveraged to deploy novel neural net models on a commercial level that can detect faults and predict remaining useful life of equipment using IoT sensors. In-house designed sensors are placed on client sites that send vibrational, temperature and humidity data to AWS DynomoDB through IoT data pipeline. Using EMR Hadoop cluster, data is transferred to S3, where it is pulled by the machine image and processed through a variety of digital preprocessing techniques. Features are extracted from the processed data before being fed into two trained neural networks. The first ANN runs diagnosis to detect the fault in the equipment while the second one predicts the Remaining Useful Life (RUL) through novel prognostic techniques. Diagnostic and prognostic metrics are deployed on dashboard that is client site specific. The dashboard can be accessed through either an online portal or iOS/Android app. The artificial neural networks were tested to have an accuracy of over 97% in predicting fault in equipment and over 90% in predicting RUL. Over 100 sensors were deployed in a successful pilot study that spanned on client sites located across the world. The current system can be scaled to over a million sensors.

Biography :



Google Scholar citation report
Citations: 2279

Journal of Computer Science & Systems Biology received 2279 citations as per Google Scholar report

Journal of Computer Science & Systems Biology peer review process verified at publons

Indexed In

arrow_upward arrow_upward