Abstract
Neural networks have been applied as a machine learning tool in many different areas. Recently, they have gained increased attention with what is now called deep learning. Neural networks algorithms have several parameters that need to be tuned in order to maximize performance. The definition of these parameters can be a difficult, extensive and time consuming task, even for expert users. One approach that has been successfully used for algorithm and parameter selection is metalearning. Metalearning consists in using machine learning algorithm on (meta)data from machine learning experiments to map the characteristics of the data with the performance of the algorithms. In this paper we study how a metalearning approach can be used to obtain a good set of parameters to learn a neural network for a given new dataset. Our results indicate that with metalearning we can successfully learn classifiers from past learning tasks that are able to define appropriate parameters.

Abstract
The reliability and safety of industrial machines depends on their timely maintenance. The integration of Cyber Physical Systems within the maintenance process enables both continuous machine monitoring and the application of advanced techniques for predictive and proactive machine maintenance. The building blocks for this revolution-embedded sensors, efficient preprocessing capabilities, ubiquitous connection to the internet, cloud-based analysis of the data, prediction algorithms, and advanced visualization methods-are already in place, but several hurdles have to be overcome to enable their application in real scenarios, namely: the integration with existing machines and existing maintenance processes. Current research and development efforts are building pilots and prototypes to demonstrate the feasibility and the merits of advanced maintenance techniques, and this paper describes a system for the industrial maintenance of sheet metal working machinery and its evolution towards a full proactive maintenance system.

Abstract
This paper proposes the linear frequency cepstral coefficients as highly discriminative features for anomaly detection in ball bearings using vibration sensor data. These features are based on cepstral analysis and are capable of encoding the patterns of a spectral magnitude profile. Incipient damages on bearings can grow rapidly under normal use resulting in vibration and harsh noise. If left undetected, this damage will worsen, leading to high maintenance costs or even injury. Multiple interferences in an industrial environment contaminate the signal, making it a challenge to correctly identify the bearings' condition. Many studies have attempted to overcome this issue at the signal level. However, the discriminative capacity of the current vibration signal features is still vulnerable to interference, which motivates this work. In order to demonstrate the benefits of these features, we (1) show that they are computationally efficient and suitable for real-time incremental training; (2) conduct discriminative analysis by evaluating the separability performance and comparing it with the state of the art; and (3) test the robustness of the proposed features under noise interference, which is ideal for use in the harsh operating conditions of industrial machinery. The data was obtained from a laboratory workbench setting that reproduces bearing fault scenarios. Results show that the proposed features are fast, competitive when compared to state-of-the-art features, and resilient to high levels of interference. Despite the higher performance when using the quadratic model, the proposed features remain highly discriminative when used with several other discriminant function.

Abstract
Sensors are susceptible to failure when exposed to extreme conditions over long periods of time. Besides they can be affected by noise or electrical interference. Models (Machine Learning or others) obtained from these faulty and noisy sensors may be less reliable. In this paper, we propose a data augmentation approach for making neural networks more robust to missing and faulty sensor data. This approach is shown to be effective in a real life industrial application that uses data of various sensors to predict the wear of an automotive fuel-system component. Empirical results show that the proposed approach leads to more robust neural network in this particular application than existing methods.

Abstract
Assembly lines are at the core of many manufacturing systems, and planning for a well-balanced flow is key to ensure long-term efficiency. However, in flexible configurations such as Multi-Manned Assembly Lines (MMAL), the balancing problem also becomes more challenging. Due to the increased relevance of these assembly lines, this work aims to investigate the MMAL balancing problem, to contribute for a more effective decision-making process. Therefore, a new approach is proposed based on Deep Reinforcement Learning (DRL) embedded in a Digital Twin architecture. The proposed approach provides a close-to-reality training environment for the agent, using Discrete Event Simulation to simulate the production system dynamics. This methodology was tested on a real-world instance with preliminary results showing that similar solutions to the ones obtained using optimization-based strategies are achieved. This research provides evidence of success in terms of dynamic resource assignment to tasks and workers as a basis for future developments.

Abstract
The ability to describe business processes as executable models has always been one of the fundamental premises of workflow management. Yet, the tacit nature of human knowledge is often an obstacle to eliciting accurate process models. On the other hand, the result of process modeling is a static plan of action, which is difficult to adapt to changing procedures or to different business goals. In this article, we attempt to address these problems by approaching workflow management with a combination of learning and planning techniques. Assuming that processes cannot be fully described at buildtime, we make use of learning techniques, namely Inductive Logic Programming (ILP), in order to discover workflow activities and to describe them as planning operators. These operators will be subsequently fed to a partial-order planner in order to find the process model as a planning solution. The continuous interplay between learning, planning and execution aims at arriving at a feasible plan by successive refinement of the operators. The approach is illustrated in two simple scenarios. Following a discussion of related work, the paper concludes by presenting the main challenges that remain to be solved.