(128a) AI for Predictive Maintenance with or without Vibration Sensing

New technologies such as AI must fit in, and add value, into legacy plant operations. During the last years many obstacles and objections have been overcome, such as:

• Putting in more sensors can be a showstopper.
• Plenty of asset health and AI/ML software demand vibration sensor data.
• Artificial Intelligence and Machine Learning applications can give insights to their users even without vibration data.
• Investment justification is possible when a logical argument can be made around the right data being available to reach an investment decision based on ROI.

This abstract covers angles regarding process industries asset health analytics and AI applications with minimal vibration data availability. It is backed by multiple installations in US facilities across different industries and focuses on the bottom-line impact of early detection of anomalies of large fleets of assets through application of priority attention index to bring items to the attention of front-line maintenance workers.

By joining this session, you will learn about trends in different industries, enabling returns from AI predictive maintenance systems, deployment across a wide range of assets and asset types, as well as easy utilization by current organization and employees of your organization. Example architecture of an AI system for predictive maintenance is evaluated from data sources to a core platform, to operational contextualization, to anomaly detection, and finally to prioritized notifications for maintenance-responsible colleagues.

Practical considerations regarding rotating equipment, fixed assets, device and sensor data for process and equipment are included, as are scaling at a specific site location and globally - with a focus on impact when it comes to sustainable operations, mobile working, related risks and costs, increased availability, and precise execution of predictive maintenance. Another practical focus of this abstract is what is possible with limited or absent vibration data.

Most maintenance practices are based on service-interval schedules, a preventative and reactive approach that doesn’t take actual machine or equipment usage and health into account, and which is questionable in substantially reducing unplanned downtime and improving maintenance effectiveness. Predictive maintenance (PdM), on the other hand, is a proactive approach, enabling machine and equipment failures to be dealt with before they stop production. PdM is achieved by analyzing huge volumes of machine and maintenance data to decode the health of machines and equipment and enable maintenance staff to optimize their activities. By predicting break downs, companies can therefore eliminate sudden failures, reduce unplanned downtime, increase machine life, optimize scheduled maintenance, and reduce the routine replacement of parts that may, in fact, be perfectly healthy.

This abstract is aimed at intermediate level participants when it comes to AI and predictive maintenance ranging from front-line workers to experts to managers and directors.