Industrial maintenance has undergone a paradigm shift due to the emergence of artificial intelligence (AI), the Internet of things, and cloud computing. Rather than accepting the drawbacks of reactive maintenance, leading firms worldwide are embracing “predict-and-prevent” maintenance. However, opaque box AI models are sophisticated and complex for the average user to comprehend and explain. This limits the AI employment in predictive maintenance, where it is vital to understand and evaluate the model before deployment. In addition, it is also important to comprehend the maintenance system's decisions. This article presents a type-2 fuzzy-based explainable AI (XAI) system for predictive maintenance within the water pumping industry. The proposed system is optimized via big-bang big-crunch, which maximizes the model accuracy for predicting faults while maximizing model interpretability. We evaluated the proposed system on water pumps using real-time data obtained by our hardware placed at real-world locations around the United Kingdom and compared our model with Type-1 fuzzy logic system (T1FLS), a multilayer perceptron (MLP) neural network, an effective deep learning method known as stacked autoencoders (SAEs), and an interpretable model such as decision trees (DT). The proposed system predicted water pumping equipment failures with good accuracy (outperforming the T1FLS accuracy by 8.9% and DT by 529.2% while providing comparable results to SAEs and MLPs) and interpretability. The system predictions comprehend why a specific problem may occur, which leads to better and more informed customer visits to reduce equipment failure disturbances. It will be shown that 80.3% of water industry specialists strongly agree with the model's explanation, determining its acceptance.