Search

Iterative learning control (ILC) can yield superior performance for repetitive tasks while only requiring approximate models, making this control strategy very appealing for industry. However, applying it to non-linear systems involves solving of optimization problems, which limits the industrial uptake, especially for learning online to compensate for variations throughout the system’s lifetime. Industry tackles this by designing simple rule-based learning controllers. However, these are often designed in an ad-hoc manner, which potentially limits performance. In this paper, we will couple a low-dimensional parametrized learning control algorithm with a generic signal parametrization method on the basis of machine learning, and specifically using autoencoders. This will allow high control performance, while limiting implementational complexity and maintaining interpretability, paving the way for a higher industrial uptake of learning control for non-linear systems. We will illustrate the parametrized approach in simulation on a non-linear slider-crank system, and provide an example of using the learning approach to perform a tracking task for this system.

In this paper the challenge of improving throughput in additive manufacturing and other laser-based manufacturing processes is addressed. It is the main obstacle, for industries using these processes, to market position themselves with regards to conventional material processing technologies. A key component affecting throughput, besides the material alteration process itself, is the laser scanner controlling the laser spot position.In this paper, a new model-based methodology to improve the performance of such laser scanners is proposed. This methodology consists of two main steps: (1) identification of a high-fidelity model of the laser scanner quantifying its dynamics (2) design of a model-based feedforward controller improving performance beyond conventional control. This methodology is implemented and validated using an industrial setup equipped with an open controller for additive manufacturing machines, an industrial laser scanner and a camera based monitoring system that allows an accurate evaluation of the scanner positioning. The validation on the setup showed improvement in performance up to 49% compared to conventional control of the laser scanner.

Dynamic models of mechatronic systems are abundantly used in the context of motion control and design of complex servo applications. In practice, these systems are often plagued by unknown interactions, which make the physics-based relations of the system dynamics only partially known. This paper presents a neural network augmented physics (NNAP) model as a combination of physics-inspired and neural layers. The neural layers are inserted in the model to compensate for the unmodeled interactions, without requiring direct measurements of these unknown phenomena. In contrast to traditional approaches, both the neural network and physical parameters are simultaneously optimized, solely by using state and control input measurements. The methodology is applied on experimental data of a slider-crank setup for which the state dependent load interactions are unknown. The NNAP model proves to be a stable and accurate modeling formalism for dynamic systems that ab initio can only be partially described by physical laws. Moreover, the results show that a recurrent implementation of the NNAP model enables improved robustness and accuracy of the system state predictions, compared to its feedforward counterpart. Besides capturing the system dynamics, the NNAP model provides a means to gain new insights by extracting the neural network from the converged NNAP model. In this way, we discovered accurate representations of the unknown spring force interaction and friction phenomena acting on the slider mechanism.