A real-time quality monitoring framework for steel friction stir welding using computational intelligence

In this article, we present a human-centric model-based framework, where we create a new process monitoring algorithm relying on discrete frequency-based analysis of process parameters. The algorithm is capable of providing real-time feedback to the process operator in linguistic form (natural language – rule base). The proposed framework is applied to the friction stir welding process, to monitor in real-time for the first time the joining of shipbuilding steel plates (DH36). We take advantage of principles of human-like information capture in granular computing (GrC) and computational intelligence (CI) to (a) build a data-driven model to predict in real-time (during welding) quantitative part quality markers extracted from frequency spectra of the process variables (downward and traverse forces), and (b) we introduce a process monitoring algorithm that takes advantage of the developed model to provide continuous feedback to the operator – in linguistic format – on the performance of the process. We conclude the study by evaluating the proposed approach based on interval type-2 radial basis function neural network (IT2-RBF-NN) against a multilayer perceptron neural network (MLP-NN), and a type-1 radial basis function neural network (T1-RBF-NN). Simulation results show the effectiveness of the proposed approach to handle uncertainties and produce reasonable process performance predictions (∼80% accuracy in testing data) that could be used to further optimise the process.