The rise of Industry 4.0 and data-intensive manufacturing makes Advanced Process Control (APC) applications more relevant than ever for process/production optimization, related costs reduction, and increased efficiency. One of the most important APC technologies is Virtual Metrology (VM). VM aims at exploiting information already available in the process/system under exam, to estimate quantities that are costly or impossible to measure. Machine Learning approaches are the foremost choice to design VM solutions. A serious drawback of traditional Machine Learning methodologies is that they require a features extraction phase that generally limits the scalability and performance of VM solutions. Particularly, in presence of multi-dimensional data, the feature extraction process is based on heuristic approaches that may capture features with poor predictive power. In this work, we exploit modern Deep Learning-based technologies that are able to automatically extract highly informative features from the data, providing more accurate and scalable VM solutions. In particular, we exploit Deep Learning architectures developed in the realm of Computer Vision to model data that have both spatial and time evolution. The proposed methodology is tested on a real industrial dataset related to Etching, one of the most important Semiconductor Manufacturing processes. The dataset at hand contains Optical Emission Spectroscopy data and it is paradigmatic of the feature extraction problem in VM under examination.

In manufacturing industries, it is of fundamental importance to detect anomalies in production in order to meet the required quality goals and to limit the number of defective products that are accidentally delivered to the customers. Nevertheless, monitoring systems currently employed in production are typically very simple and rely on a set of univariate control charts that fail to capture the multivariate and complex nature of real-world industrial systems. In such context, Machine Learning (ML)-based approaches for Anomaly Detection (AD) have proven to be extremely effective in increasing anomalies detectability and, in general, in enhancing monitoring procedures. However, industrial data are typically very complex and not suitable to be fed directly to classical ML-based AD tools making feature extraction procedures a necessary step that unfortunately may lead to information loss and low scalability. Deep Learning, has proven very effective at learning useful representations of complex data in an automatic way. In this paper, we propose an AD pipeline that makes use of convolutional autoencoders to extract useful features from two-dimensional, non-image, data. We test our approach on real world Optical Emission Spectroscopy data that are typical of semiconductor manufacturing and we achieve improved performance over classical monitoring methods. Note to Practitioners —Advanced monitoring is one of the most important task in the context of Industry 4.0. Some of the main issues in developing Machine Learning-based solutions in industrial environment are: (i) the lack of reliable tagged data; (ii) the complexity of data structures present in real-world scenarios. In this paper we investigate unsupervised anomaly detection for 2-dimensional data in manufacturing environment: we provide an approach that exploit Deep Learning-based architecture for handling the data at hand. We show the effectiveness of the proposed approach in a real world case study related to optical emission spectroscopy data in semiconductor manufacturing process providing satisfactory classification accuracy.

Several faults affect heating, ventilation, and air conditioning (HVAC) chiller systems, leading to energy wastage, discomfort for the users, shorter equipment life, and system unreliability. Early detection of anomalies can prevent further deterioration of the chiller and energy wastage. In this work, a data-driven approach is used in order to detect faults that usually plague chiller systems. In particular, the proposed approach employs a kernel principal component analysis (KPCA) in order to capture the normal operative conditions of the system; the learning method turns out to be effective in handling nonlinear phenomena through the use of the Gaussian kernel, which, by means of a self-tuning procedure, ensures good accuracy properties while maintaining enough generalization characteristics. The effectiveness of the proposed fault detection method is evaluated by means of tests on emulated and real chiller data. The KPCA approach is first proved to exhibit better detection performances than linear PCA, and then, it is corroborated through the comparison with local outlier factor, one-class support vector machine, and isolation forest.

Predictive Maintenance technologies are particularly appealing for Industrial Equipment producers, as they pave the way to the selling of high added-value services and customized maintenance plans. However, standard Predictive Maintenance approaches assume the availability of sensor measurements, and the costs associated with adding sensors or remotely accessing sensor readings may discourage the development of such technologies. In this context, Alarm Forecasting can be very useful as it represents a low-cost alternative or helpful support to sensor-based Predictive Maintenance. In this work, we propose a new formulation for the Alarm Forecasting problem, framed as a multi-label classification task.