Machine Learning-Based Predictive Maintenance: Using CNN – LSTM network

Maintenance is one of the critical operations undertaken in any manufacturing facility; hence, it represents an essential function for every plant. However, it is considered a non-added value to the product that demands minimizing its cost function. With maintenance costing around 15% to 60% of the total plant conversion budget, the modern factory must employ effective and smart predictive maintenance techniques. This is where the potential of machine learning powered by predictive algorithms can provide an edge on both the cost and the availability of the machines and the plant. This paper discusses predictive maintenance, machine learning models, and techniques. It develops a CNN-LSTM architecture that facilitates learning from collected data from a selected system and successfully conducts its predictive maintenance functions. The developed Machine Learning-based prediction algorithm is tested using potential datasets with different applications to establish a benchmark for its performance and allow its benefits to be measured and analyzed. Also, the paper compares the obtained results with those that represent the state-of-the-art relevant to different learning algorithms available in the literature and suggests future work for potential extension and performance improvement.