New algorithm combines existing approaches but offers 100 times better results

Fault detection 100 times faster with new technique

British engineers have created a more powerful algorithm to detect faults in industrial machines and wind turbines.

The team from the University of Sheffield and the University of Lincoln has combined two existing approaches to achieve much faster detection and classification of faults in mechanical processes, promising to cut down false alarms and increase the time the machines are working.

The two methods utilised in the new algorithm, are the real-coded genetic algorithm and the K-means clustering methodology.

While separately, each of these methods presents considerable shortcomings, combining them into a new tool named G3Kmeans helps to harness the best of each and combine their strengths.

“Data from industrial machines often involves a very complex search space,” said Dr Jun Chen, a researcher at Lincoln University. “If you use conventional clustering algorithms, you end up with misclassification. For example, non-faults are flagged as faults and vice versa. With the G3Kmeans algorithm you can reach a reliable classification – that is the first step towards optimal maintenance,” he said.

K-means clustering uses a group of set objects and groups those objects that are more similar to each other into clusters. Each observation then belongs to the cluster with the nearest central value.

The genetic algorithm, on the other hand, proposes a number of possible solutions to a given problem, aiming to achieve a better solution. Pre-selected properties of each solution can be altered in subsequent reiterations until the optimum answer is found. However, the process is rather lengthy requiring sometimes more than 1000 reiterations.

“You run the algorithm many times and every time the algorithm is modified based on the solution found from the last one, gradually improving the solution to the problem each time,” said Chen.

The improved G3Kmeans algorithm, on the contrary, can find the optimum solution and detect the fault in only 11 repetitions.

“In terms of industrial machines this is the first step in creating an algorithm that optimises the search for a solution and can identify a fault with certainty,” said Chen. ”This method will reduce maintenance costs by reducing the amount of false alarms requiring investigation.”

The team is now looking to optimise the strategy by making the algorithm more wide-ranging to enable its use to detect solutions for a variety of specific applications.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them