The idea is to analyze historical data patterns that are leading to device failures, find the leading indicators, and start using those patterns to predict failure before it happens again on the same device or other similar devices. The more data we have from similar devices or assets, the more accurate the prediction algorithms are
going to be.

The initial steps for prescriptive analytics, are systems that use artificial intelligence and expert system techniques to continuously learn from human maintenance prescriptions under different circumstances, to become an automated “second opinion” for the human that proposes several options based on analyzed past decisions. The next prescriptive level would be for the machine to start making recommendations for actions directly, with high confidence levels, from analyzing the input data.