AI Predictive Maintenance

Over the years the increasing cost of capital equipment and stress on the supply chains of most companies have increased the impact of single point failures. Lean Six Sigma (LSS) historically relied on Total Preventive Maintenance (TPM), as described in 1984 by Seiichi Nakajima 28 years before Deep Neural Networks became a practical tool. The TPM process was indeed a great step forward, but it ultimately concluded with a regular fixed schedule of maintenance for each machine. This approach has come under due criticism as it can result in machine shutdowns when either maintenance is in fact not needed or when machines fail in between scheduled maintenance causing lost production.

With the advent of AI and ever-improving sensors it has become possible to monitor machine performance and actually predict when failures may occur.  Employing AI Accelerator and our deep process knowledge we’re able to approach zero downtime and enable greater business and mission success leveraging dynamic, real time data streaming from shop floor systems.  Identifying hidden patterns of unusual or degrading performance in advance of catastrophic and costly failures we’re able to take profit from smart investments in prevention (planned maintenance), which are much easier to manage and far less costly than correction of failures (unplanned maintenance).

One of the most common failures in Petrochemical, Refining, and Machining etc. involves bearing wear.  With real time monitoring of noise amplitudes, frequencies and harmonics, data captured can be used to train an AI Neural Network, which can continuously assess and determine the condition of bearings and other wear points within shop floor systems.  Equipped with this greatly improved, real time insight into predicted failure rates, Production management can order spare parts just-in-time and schedule maintenance personnel accordingly (see Figure 12-3).

12-3 Average Amplitude of the Spall Frequency

An added advantage, when a machine is shut down prior to failure, is that the “suspension mode” allows a better estimate of residual life of the equipment.  The ultimate goal is to eliminate unplanned and unscheduled downtime for or maintenance, thus protecting and enhancing overall equipment effectiveness and utilization.

References:

“Lean Six Sigma in the Age of Artificial Intelligence”, Chapter 12

https://sites.ualberta.ca/~ztian/index_files/Papers/MSSP_2010.pdf