This study investigates the predictive modelling of weld bead geometry in wire arc additive manufacturing (WAAM) through advanced machine learning methods. While WAAM is valued for its ability to produce large, complex metal parts with high deposition rates, precise control of the weld bead remains a critical challenge due to its influence on mechanical properties and dimensional accuracy. To address this problem, this study utilized machine learning approaches—Ridge regression, Lasso regression and Bayesian ridge regression, Random Forest and XGBoost—to predict the key weld bead characteristics, namely height, width and cross-sectional area. A Design of experiments (DOE) was used to systematically vary the welding current and travelling speed, with 3D weld bead geometries captured by laser scanning. Robust data pre-processing, including outlier detection and feature engineering, improved modelling accuracy. Among the models tested, XGBoost provided the highest prediction accuracy, emphasizing its potential for real-time control of WAAM processes. Overall, this study presents a comprehensive framework for predictive modelling and provides valuable insights for process optimization and the further development of intelligent manufacturing systems.

This study applied a holistic approach to the problem of controlling the temperature of critical areas of tools using conformal cooling. The entire injection molding process is evaluated at the tool design stage using four criteria, one from each stage of the process cycle, to produce a tool with effective cooling that enables short cycle times and ensures good product quality. Tool manufacturing time and cost, as well as tool life, are considered in the optimization by introducing a novel tool-efficiency index. The multi-objective optimization is based on numerical simulations. The simulation results show that conformal cooling effectively cools the critical area of the tool and provides the shortest cycle times and the lowest warpage, but this comes with a trade-off in the tool-efficiency index. By using the tool-efficiency index with non-dominated sorting, the number of relevant simulation cases could be reduced to six, which greatly simplifies the decision regarding the choice of cooling system and process parameters. Based on the study, a tool with conformal cooling channels was made, and a coolant inlet temperature of 20 °C and a flow rate of 5 L/min for conformal and 7.5–9.5 L/min for conventional cooling channels were selected for production. The simulation results were validated by experimental measurements.

Successful control of a dielectric elastomer actuator (DEA) can be a challenging task, especially if no overshoot is desired. The work presents the first use of the $P{I}^{\lambda }{D}^{\mu }$ control for a dielectric elastomer actuator to eliminate the overshoot. The mathematical model of the dielectric elastomer was established using the fractional Kelvin-Voigt model. Step responses are first tested in the Laplace domain, which gave the most satisfactory results. However, they did not represent the real model. It cannot have negative force acting on the dielectric elastomer actuator. Simulations in Matlab/Simulink were performed to obtain more realistic responses, where output of the $P{I}^{\lambda }{D}^{\mu }$ controller was limited. Initial parameters for a PID control were obtained by the Wang–Juang–Chan algorithm for the first order plus death time function approximation to the step response of the model, and reused as the basis for the $P{I}^{\lambda }{D}^{\mu }$ actuator control. A quasi-anti-windup method was introduced to the final control algorithm. Step responses of the PID and the $P{I}^{\lambda }{D}^{\mu }$ in different domains were verified by simulation and validated by experiments. Experiments proved that the fractional calculus $P{I}^{\lambda }{D}^{\mu }$ step responses exceeded performance of the basic PID controller for DEA in terms of response time, settling time, and overshoot elimination.

Actuators made of dielectric elastomers are used in soft robotics for a variety of applications. However, due to their mechanical properties, they exhibit viscoelastic behaviour, especially in the initial phase of their performance, which can be observed in the first cycles of dynamic excitation. A fully fractional generalised Maxwell model was derived and used for the first time to capture the viscoelastic effect of dielectric elastomer actuators. The Laplace transform was used to derive the fully fractional generalised Maxwell model. The Laplace transform has proven to be very useful and practical in deriving fractional viscoelastic constitutive models. Using the global optimisation procedure called Pattern Search, the optimal parameters, as well as the number of branches of the fully fractional generalised Maxwell model, were derived from the experimental results. For the fully fractional generalised Maxwell model, the optimal number of branches was determined considering the derivation order of each element of the branch. The derived model can readily be implemented in the simulation of a dielectric elastomer actuator control, and can also easily be used for different viscoelastic materials.