A recommender system for human operators in industrial automation

Abstract

Listen

Translate article

This paper presents a recommender system to assist human operators in industrial automation applications, which suggests proper actions to take during process operation and those to avoid. The proposed recommender system is built based on a three-stage computational procedure. In the first stage, historical process data segments containing timestamped events are clustered using a k-medoids-based algorithm, where instead of distance metrics, structural similarity of the segments is used. This is achieved by using a Smith–Waterman-based algorithm to consider all events and their timestamps in the segments. In the second stage, a modified collaborative filtering technique applicable to time sequences is proposed, where not only operator ratings, but also the structural similarity of the segments are taken into account. In the third stage, possible actions are analyzed and accordingly, proper and improper actions are determined and presented to the human operators. The effectiveness of the proposed recommender system is examined through a power system operation example.