Maintenance and Artificial Intelligence

Let's talk about Maintenance and Artificial Intelligence

Although Machine Learning offers us great possibilities in everything related to maintenance, we are going to talk about other strategies.
Imagine that you want to implement an advanced fault diagnosis system for a wind turbine using Fuzzy Logic and Deep Learning techniques in order to obtain reliability (dual fault diagnosis system).

How could it be done?

With NNs:
After data collection (Vibration, Temperature, Pressure, etc), labeling and training, you can use deep learning to automatically classify patterns into different fault categories.

With Fuzzy Logic:
Fuzzy logic can be used to analyze data from sensors installed on turbine components. You could then define the fuzzy sets for the levels of the different parameters, such as "Low", "Moderate", and "High", based on historical data and expert knowledge.

A set of fuzzy rules must be established that relate the levels of the different parameters with potential failures. (e.g : IF vibration is high in the rear gearbox bearing AND vibration is moderate in the front bearing of generator THEN Possible gearbox misalignment.)
After this, the inference engine could be defined in the fuzzy logic system, it processes the fuzzy rules to determine the degree to which each rule is applied based on the current paremeter levels.

Finally the fuzzy logic system processes these rules to estimate the possibility of specific operational issues through "defuzzification" process, give us responses about potential failures in the system.
In this way we would then have a redundant system that would allow us to make maintenance decisions in a better way.