In industrial applications, the continuously growing development of multi-sensor approaches, together with the trend of creating data-rich environments, are straining the effectiveness of the traditional Statistical Process Control (SPC) tools. Industrial data streams frequently violate the statistical assumptions on which SPC tools are based, presenting non-normal or even mixture distributions, strong autocorrelation and complex noise patterns.\ud \ud To tackle these challenges, novel nonparametric approaches are required. Machine learning techniques are suitable to deal with distributional assumption violations and to cope with complex data patterns. Recent studies showed that those methods can be used in quality control problems by exploiting only in-control data for training (such a learning paradigm is also known as “one-class-classification”).\ud \ud In recent studies, the use of distribution-free multivariate SPC methods was proposed, based on unsupervised statistical learning tools, pointing out the difficulty of defining suitable control regions for non-normal data. In this paper, a Self-Organizing Map (SOM) based monitoring approach is presented. The SOM is an automatic data-analysis method, widely applied in recent works to clustering and data exploration problems. A very interesting feature of this method consists of its capability of providing a computationally efficient way to estimate a data-adaptive control region, even in the presence of high dimensional problems. Nevertheless, very few authors adopted the SOM in an SPC monitoring strategy. The aim of this work is to exploit the SOM network architecture, and proposing a network design approach that suites the SPC needs. A comparison study is presented, in which the process monitoring performances are compared against literature benchmark methods. The comparison framework is based on both simulated data and real data from a roll grinding application

The machine tool industry is facing the need to increase the sensorization of production systems to meet evolving market demands. This leads to the increasing interest for in-process monitoring tools that allow a fast detection of faults and unnatural process behaviours during the process itself. Nevertheless, the analysis of sensor signals implies several challenges. One major challenge consists of the complexity of signal patterns, which often exhibit a multiscale content, i.e., a superimposition of both stationary and non-stationary fluctuations on different time-frequency levels. Among time-frequency techniques, Empirical Mode Decomposition (EMD) is a powerful method to decompose any signal into its embedded oscillatory modes in a fully data-driven way, without any ex-ante basis selection. Because of this, it might be used effectively for automated monitoring and diagnosis of manufacturing processes. Unfortunately, it usually yields an over-decomposition, with single oscillation modes that can be split into more than one scale (this effect is also known as “mode mixing”). The literature lacks effective strategies to automatically synthetize the decomposition into a minimal number of physically relevant and interpretable components. This paper proposes a novel approach to achieve a synthetic decomposition of complex signals through the EMD procedure. A new criterion is proposed to group together multiple components associated to a common time-frequency pattern, aimed at summarizing the information content into a minimal number of modes, which may be easier to interpret. A real case study in waterjet cutting is presented, to demonstrate the benefits and the critical issues of the proposed approach

The quality assessment of manufacturing processes has been traditionally based on sample measures performed on the process output. This leads to the common “product-based Statistical Process Control (SPC)” framework. However, there are applications of actual industrial interest where post-process quality measurement procedures involve time-consuming inspections strongly related to the operator’s experience and/or based on expensive equipment. Cylindrical grinding of large rolls may be one of them. The assessment of the final acceptability of a ground cylinder, in terms of surface finish, is a challenging task with traditional measuring tools, and it often depends on operator’s visual inspections and on his subjective evaluations. In this frame, a paradigm shift is required to substitute troublesome post-process monitoring procedures with in-process and signal-based ones. The paper reviews the quality control issues in surface quality monitoring of big ground rolls where process vibrations (i.e. chatter) are one of major concerns. A multi-sensor approach for vibration onset detection, based on the use of a multi-channel implementation of the Principal Component Analysis, is proposed. The potential benefits, the implementation issues, and the main criticalities are discussed by analysing data of a real industrial application.