Nowadays, artificial intelligence (AI) has become a crucial Key Enabling Technology with extensive application in diverse industrial sectors. Recently, considerable focus has been directed towards utilizing AI for the development of optimal control in industrial processes. In particular, reinforcement learning (RL) techniques have made significant advancements, enabling their application to data-driven problem-solving for the control of complex systems. Since industrial manufacturing processes can be treated as MIMO non-linear systems, RL can be used to develop complex data-driven intelligent decision-making or control systems. In this work, the workflow for developing a RL application for industrial manufacturing processes, including reward function setup, development of reduced order models and control policy construction, is addressed, and a new process-based reward function is proposed. To showcase the proposed approach, a case study is developed with reference to a wire arc additive manufacturing (WAAM) process. Based on experimental tests, a Reduced Order Model of the system is obtained and a Deep Deterministic Policy Gradient Controller is trained with aim to produce a simple geometry. Particular attention is given to the sim-to-real process by developing a WAAM simulator which allows to simulate the process in a realistic environment and to generate the code to be deployed on the motion platform controller.

This review article delves into the conceptual framework of digital twins and their diverse applications across research domains, highlighting the pivotal role of machine learning in shaping the development and integration of digital twin technology across multiple disciplines. Emphasising key features like multidisciplinarity and multi-scale aspects, the paper explores how data-driven techniques are employed for modelling, visualisation, monitoring, and optimisation within the digital twin framework, pinpointing the benefits introduced in the current state-of-the-art applications, and elucidates persisting challenges across various research fields, including advanced materials, smart buildings, and manufacturing systems.

The study aimed to assess the performance of several unsupervised machine learning (ML) techniques in online anomaly (The term “anomaly” is used here to indicate a departure from expected process behavior which may indicate a quality issue which requires further investigation. The term “defect detection” has often been used previously but the specific imperfection is often indirectly inferred.) detection during surface tension transfer (STT)-based wire arc additive manufacturing. Recent advancements in quality monitoring for wire arc manufacturing were reviewed, followed by a comparison of unsupervised ML techniques using welding current and welding voltage data collected during a defect-free deposition process. Both time domain and frequency domain feature extraction techniques were applied and compared. Three analysis methodologies were adopted: ML algorithms such as isolation forest, local outlier factor, and one-class support vector machine. The results highlight that incorporating frequency analysis, such as fast Fourier transform (FFT) and discrete wavelet transform (DWT), for feature extraction based on general frequency response and defined bandwidth frequency response, significantly improves performance, reflected in a 14% increase in F2 score, compared with time-domain features extraction. Additionally, a deep learning approach employing a convolutional autoencoder (CAE) demonstrated superior performance by processing time-frequency domain data stored as spectrograms obtained through short-time Fourier transform (STFT) analysis. The CAE method outperformed frequency domain analysis and traditional ML approaches, achieving an additional 5% improvement in F2-score. Notably, the F2-score (The F2 score is the weighted harmonic mean of the precision and recall (given a threshold value). Unlike the F1 score, which gives equal weight to precision and recall, the F2 score gives more weight to recall than to precision.) increased significantly from 0.78 in time domain analysis to 0.895 in time-frequency analysis. The study emphasizes the potential of utilizing low-cost sensors to develop anomaly detection modules with enhanced accuracy. These findings underscore the importance of incorporating advanced data processing techniques in wire arc additive manufacturing for improved quality control and process optimization.

Welding is one of the most complex industrial processes because it is challenging to model, control, and inspect. In particular, the quality inspection process is critical because it is a complex and time-consuming activity. This research aims to propose a system of online inspection of the quality of the welded items with gas metal arc welding (GMAW) technology through the use of neural networks to speed up the inspection process. In particular, following experimental tests, the deviations of the welding parameters—such as current, voltage, and welding speed—from the Welding Procedure Specification was used to train a fully connected deep neural network, once labels have been obtained for each weld seam of a multi-pass welding procedure through non-destructive testing, which made it possible to find a correspondence between welding defects (e.g., porosity, lack of penetrations, etc.) and process parameters. The final results have shown an accuracy greater than 93% in defects classification and an inference time of less than 150 ms, which allow us to use this method for real-time purposes. Furthermore in this work networks were trained to reach a smaller false positive rate for the classification task on test data, to reduce the presence of faulty parts among non-defective parts.

Nowadays, sustainability of manufacturing processes is a major concern which calls for special efforts to reduce their environmental impact and energy consumption. In additive manufacturing, this issue is even more challenging due to the usually high energy demands of these processes. However, in the era of Industry 4.0, machine learning (ML) techniques, combined with metaheuristic optimization algorithms, offer a powerful solution to explore new, unproven combinations of process parameters that better align with sustainability goals of manufacturing. These methodologies can minimize the need for extensive experimental campaigns and provide a valuable decision-making support tool for goal-oriented process parameters optimization. In line with such approach, this research work introduces an intelligent data-driven methodology using ML to optimize wire arc additive manufacturing (WAAM) of Invar 36 alloy considering both the resulting layer geometry and quality as well as the process energy consumption. An experimental campaign involving WAAM deposition of 15 walls made of Invar 36 alloy using a natural dip transfer welding process was carried out. The data acquired from the WAAM experimental tests were used to develop and train an artificial neural network (ANN) which, on the basis of the process parameters, was able to predict the layer geometry, the specific energy consumption and a specified quality score indicative of the presence of defects. The ANN achieved a high accuracy with 100% F2 score for quality classification, 0.4 mm mean absolute error for layer geometry, and 20 J/mm MAE for specific energy consumption. A genetic algorithm (GA) was then used to identify optimal process parameters able to minimize the specific energy consumption while maintaining quality and smoothness of the deposited layer. The experimental validation carried out using the GA-optimized process parameters in the WAAM process confirmed the reliability of the model, resulting in energy-efficient and defect-free walls.