Why it is so important to accompany Artificial Intelligence with the so called “Knowledge Domain” (KD) in industrial applications?
Knowledge Domain is defined as a broad-based understanding of a particular body of information (Encyclopedia of Science and Technology). It allows effective interpretation of data correlations, events or other symptoms in a related process.
Industrial applications of AI frequently find relevant correlations that appear strange and difficult to understand. Other correlations are evident and offer no additional information when investigating a failure mechanism, whether it is for predicting or root cause elimination.
Understanding these strange correlations requires process knowledge (domain) to realize their true meaning in terms of hidden root causes. In most cases these unveil Systemic Root Causes, that when corrected, produce dramatic improvements.
One example to illustrate the point:
The position of a backpressure control valve was reported, by the analytics module, to have a strong correlation with chronic breaking and pitting of heat exchanger plates in a cooling system of an offshore platform, after looking in detail into the valve operation and CMMS records, among others, it became clear that cavitation and associated vibrations were affecting the control valve, heat exchanger, isolation valves, as well as piping. Reducing water velocity proved to be very effective in eliminating several failure mechanisms linked to same systemic root cause.
Assembling the required holistic knowledge domain (team) for root cause failure analysis powered with AI is a new role for reliability engineers and organizations.