Surface roughness prediction in milling using long-short term memory modelling

Abstract

Artificial Intelligence (AI), presently vogue in technology, energizes many researchers to address complex issues. As a result of advances in machine learning and data analytics, the manufacturing cycle has become more efficient. Optimizing machining parameters is vital to ensure better surface quality. To enable proactive actions in manufacturing, incipient surface roughness prediction is quite essential. The challenge of capturing non-linear dynamics is becoming more complex as the data grows. Traditional machine learning techniques are unable to express the sequential features extracted. The recent approach, Long Short-Term Memory (LSTM), can handle a variety of data lengths and extract long-term series features. In this paper, the LSTM approach is utilized to forecast surface roughness during milling of the S45C steel dataset. On the test data set, the Root Mean Square Error (RMSE) loss is computed using the LSTM model, yielding an RMSE loss of 0.1097. In data-driven smart manufacturing, the LSTM model demonstrates its capabilities in surface roughness decision-making.