SCOPUS eid=2-s2.0-74549178470 - The quality of products and processes is more and more often becoming related to functional data, which refer to information summarised in the form of profiles. The recent literature has pointed out that traditional control charting methods cannot be directly applied in these cases and new approaches for profile monitoring are required. While many different profile monitoring approaches have been proposed in the scientific literature, few comparison studies are available. This paper aims at filling this gap by comparing three representative profile monitoring approaches in different production scenarios. The performance comparison will allow us to select a specific approach in a given situation. The competitor approaches are chosen to represent different levels of complexity, as well as different types of modelling approaches. In particular, at a lower level of complexity, the 'location control chart' (where the upper and lower control limits are +/-K standard deviations from the sample mean at each profile location) is considered to be representative of industrial practice. At a higher complexity level, approaches based on combining a parametric model of functional data with multivariate and univariate control charting are considered. Within this second class, we analyse two different approaches. The first is based on regression and the second focuses on using principal component analysis for modelling functional data. A manufacturing reference case study is used throughout the paper, namely profiles measured on machined items subject to geometrical specification (roundness).

The data-rich environments of industrial applications lead to large amounts of correlated quality characteristics that are monitored using Multivariate Statistical Process Control (MSPC) tools. These variables usually represent heterogeneous quantities that originate from one or multiple sensors and are acquired with different sampling parameters. In this framework, any assumptions relative to the underlying statistical distribution may not be appropriate, and conventional MSPC methods may deliver unacceptable performances. In addition, in many practical applications, the process switches from one operating mode to a different one, leading to a stream of multimode data. Various nonparametric approaches have been proposed for the design of multivariate control charts, but the monitoring of multimode processes remains a challenge for most of them. In this study, we investigate the use of distribution-free MSPC methods based on statistical learning tools. In this work, we compared the kernel distance-based control chart (K-chart) based on a one-class-classification variant of support vector machines and a fuzzy neural network method based on the adaptive resonance theory. The performances of the two methods were evaluated using both Monte Carlo simulations and real industrial data. The simulated scenarios include different types of out-of-control conditions to highlight the advantages and disadvantages of the two methods. Real data acquired during a roll grinding process provide a framework for the assessment of the practical applicability of these methods in multimode industrial applications.

Profile monitoring can be effectively adopted to detect unnatural behaviors of machining processes, i.e., to signal when the functional relationship used to model the geometric feature monitored changes with time. Most of the literature concerned with profile monitoring deals with the issue of model identification for the functional relationship of interest, as well as with control charting of the model parameters. In this chapter, a different approach is presented for profile monitoring, with a focus on quality monitoring of geometric tolerances. This approach does not require an analytical model for the statistical description of profiles considered, and it does not involve a control charting method. An algorithm which allows a computer to automatically learn from data the relationship to represent profiles in space is described. The proposed algorithm is usually referred to as a neural network and the data set, from which the relationship is learned, consists just of profiles representative of the process in its in-control state. Throughout this chapter, a test case related to roundness profiles obtained by turning and described in Chapter 11 is used as a reference. A verification study on the efficacy of the neural network shows that this approach may outperform the usual control charting method.

Recently, non-contact sensor technologies are more and more often used for quality inspection tasks as well as for process monitoring in manufacturing. As a matter of fact, recent advancements in laser scanners and machine vision systems provide the potential to greatly improve the performance of Statistical Process Control (SPC) approaches. In this paper, a PC- based machine vision system, which provides rapid measurement of freeform geometric features, is presented. The measuring system is based on appropriate hardware and software modules. The hardware module consists of a laser scanning device and setup fixtures that can provide proper location and orientation for the part to be measured. The software module generates optimal scan plans so that the scanning operation can be performed accordingly. Furthermore, measurements for each geometric feature are automatically stored by the software module in order to perform on-line statistical analysis. The system described in this paper has been designed for on-line data acquisition, quality inspection, and statistical monitoring of actual manufacturing processes. To these aims, a user-friendly, menu-driven graphical interface has been implemented in order to give the operator an effective overview of the process state (either in-control or out-of-control). A real case study, related to the production of stamped metal panels in the automotive industry, is described.

In the last years, we are assisting to a continuous drive toward the use of hybrid metrology systems, which can combine noncontact and contact sensors to take advantage of the speed of 3-D optical sensors and of the accuracy of traditional contact solutions. We show how approaches for multisensor data fusion can be combined to control charting in order to detect (and distinguish between) out-of- control states due to the measurement and/or to the manufacturing processes.

SCOPUS eid=2-s2.0-84857541892 - This paper investigates the effect of process parameters on the shape of three-dimensional (3D) curves. Compared with recent literature on profile modeling (generally used for process monitoring applications), the main novelty consists of dealing with curves that do not lie in a plane but in space. In particular, a generalization of principal component analysis (PCA), based on the appropriate use of complex rather than real numbers (complex PCA), is used as a modeling tool. Scores of the significant components obtained via complex PCA are then used as data input for the analysis of variance to investigate the effect of process parameters on the machined shape. The first advantage of complex PCA is its insensitivity to rotation about the origin of the complex plane, a property that produces a significant simplification of the preliminary step of profile registration. Second, complex PCA outperforms alternative PCA-based approaches in terms of the power of detecting the effect of controllable factors on the profile shape. The proposed procedure is particularly useful when the quality characteristic of interest is related to geometric tolerances, as shown in a real case study of the axial straightness of lathe-turned items.

The use of neural networks began to be applied because the traditional control charts used for monitoring manufacturing process, in some cases, did not provide the possibility of correctly and quickly signalling the existing causes of variation. In today's manufacturing environment, neural networks present increasing usefulness for implementing the automation of statistical process control. This chapter targets issues on the use of neural networks for quality control of manufacturing processes, concerning the way of operation of each network model, the network's architecture and the results provided. Applications of neural networks for pattern recognition and for detection of mean and/or variance shifts in process are discussed. Comparisons between the performances of the neural approach and those of traditional control charts are also presented. Results prove that the neural network model is a useful alternative to the existing control schemes.

During the past few years, an increasing number of approaches and applications of profile monitoring have been proposed in the literature as the quality of product and process is very often characterized by functional data. In the context of geometric tolerances, where curves and surfaces describe the shape of manufactured item, the quality outcome (dependent variable) is a function of one or more spatial location variables (independent variables). Up to now, profile monitoring has been mainly constrained to situations in which the dependent variable is a scalar, which is modeled as a function of a single location variable via linear models or datareduction approaches as Principal Component Analysis (PCA). When the quality of products is related to geometric tolerances (e.g., roundness or circularity, straightness, cylindricity, flatness or planarity) the geometry of the item lies in a 3-dimensional (3D) space and cannot be modeled as a scalar function of one location variable. This paper presents solutions to problems arising when 3D features (either curves or surfaces) are considered and data-reduction techniques are implemented as modeling tool. Two PCAbased approaches are presented, namely (i) the complex PCA (i.e., PCA performed on matrices of complex numbers) and the (ii) multilinear PCA (i.e., PCA performed on tensor data). These two approaches are explored as viable solutions to modeling 3D profiles and surfaces respectively, in the context of geometric tolerance monitoring.

There has been extensive research on analysis of nonlinear profiles. Nevertheless, most research deals only with single profiles. In some industrial practices, however, the sensing system records more than one profile at each operation cycle. In this work, for the purpose of fault detection and diagnosis, we propose a method for analyzing multichannel profiles based on uncorrelated multilinear PCA. We show the effectiveness of the proposed method using simulation and a case study.

Quality of machined products is often related to the final shapes of the manufactured surfaces. This paper presents a novel method for surface monitoring, which combines Gaussian processes to model the manufactured shape and multivariate control charting for monitoring the deviations of the actual surface from the in-control pattern. Regardless of the specific case study, the proposed approach is general and can be extended to deal with different kinds of surfaces or profiles.

In manufacturing, Machine Vision Systems are increasingly being used for their ability to collect information pertaining to the quality of a product in real time. When physical profiles are collected from images captured by Machine Vision Systems, the number and locations of the observed points can change from one item to another. The research is focused on non-parametric control charts for statistical monitoring of free-form profiles with different number and locations of observed points. The proposed method consists in extracting the shape of the monitored profile from images and in comparing it to a baseline model taken as reference. A new discrepancy metric, which consists in computing the deviation area of the monitored profile from the baseline model, is proposed. Two control charting procedures, based on univariate and a multivariate statistics are illustrated and validated through computer simulations. The automatic cutting process in leather part manufacturing (e.g. furniture, automotive interior, apparel) is the reference context.

Profile monitoring has been recently considered as one of the most promising areas of research in statistical process control. Often, when physical profiles are acquired by machine vision systems, the number and locations of the argument variables can change from one item to another. In this article, non-parametric approaches for statistical monitoring of free-form profiles, e.g., characterized by irregular shapes and with different argument variables, are proposed. A machine vision system is used for process control by implementing on a computer the following procedure: (i) a feature extraction phase, which extracts contour of the produced part from image and returns it in form of polygonal curve; (ii) a registration phase, for aligning the current polygonal curve to the baseline model; (iii) a matching computation phase, which is based on the deviation area, for quantifying the discrepancies of the current polygonal curve from the baseline one; (iv) a control charting phase, which is based on univariate or multivariate statistics, for process monitoring. The automatic cutting process in leather part manufacturing is used as the reference industrial case. The leather hide is cut into parts that may assume shapes different from the ‘baseline’, e.g., the reference ideal profile. Two control charting procedures are illustrated and validated through simulated test cases and a real-life case of industrial relevance.

A relatively new area of research in the field of statistical process control has been named profile monitoring. It includes a collection of methods and techniques used to check the stability of a functional relationship over time. This chapter shows how this approach can be usefully considered as a viable solution to form error monitoring when geometric tolerances are of interest. In this case, quality monitoring consists in detecting deviations of the shape from its nominal or incontrol state. This task is accomplished by firstly modeling the functional relationship representing the manufactured shape and then checking whether or not the estimated model is stable over time. The goal of this chapter is to introduce profile monitoring, show how it works, and then illustrate how this approach can be effectively used for quality control of geometric form errors.

In modern manufacturing systems, online sensing is being increasingly used for process monitoring and fault diagnosis. In many practical situations, the output of the sensing system is represented by time-ordered data known as profiles or waveform signals. Most of the work reported in the literature has dealt with cases in which the production process is characterized by single profiles. In some industrial practices, however, the online sensing system is designed so that it records more than one profile at each operation cycle. For example, in multi-operation forging processes with transfer or progressive dies, four sensors are used to measure the tonnage force exerted on dies. To effectively analyze multichannel profiles, it is crucial to develop a method that considers the interrelationships between different profile channels. A method for analyzing multichannel profiles based on uncorrelated multilinear principal component analysis is proposed in this article for the purpose of characterizing process variations, fault detection, and fault diagnosis. The effectiveness of the proposed method is demonstrated by using simulations and a case study on a multi-operation forging process.

SCOPUS eid=2-s2.0-79952627449 - In modern manufacturing, approaches for profile monitoring can be adopted to detect unnatural behaviors of production processes, i.e. to signal when the relationship used to represent the profiles changes with time. Most of the literature concerned with profile monitoring deals with the problem of model identification and multivariate charting of parameters vector. In this paper, a different approach, which is based on an unsupervised neural network, is presented for profile monitoring. The neural network allows a computer to automatically learn from data the relationship co represent in-control profiles. Then, the algorithm may produce a signal when an input profile does not fit to the prototype learned from the in-control ones. The neural network does not require an analytical model for the statistical description of profiles faced (model-free approach). A comparison study is provided in this paper, in which the Phase II performance of the neural network is compared to that of approaches representative of the industrial practice. Performance is assessed by computer simulation, with reference to a case study related to profiles measured on machined items subject to geometrical specification (roundness). The results indicate that the neural network may outperform usual control charts in signaling out-of-control conditions, due to spindle-motion errors in several production scenarios. The proposed approach can be considered a valuable option for profile monitoring in industrial applications. (C) 2011 Elsevier Ltd. All rights reserved.

In modern manufacturing systems, online sensing has been increasingly used for process monitoring and fault diagnosis. In many practical situations, the output of the sensing system are represented by time ordered data know as “profiles” and “waveform signals”. Most of previous works dealt with cases, in which the production process was characterized by single profiles. In some industrial practices, however, the online sensing system is so designed that it records more than one profile at each operation cycle. For example, in multi-operation forging processes with transfer or progressive dies, in order to measure the tonnage force exerted on dies, four sensors are used. To effectively analyze multi-channel profiles, it is crucial to develop a method that considers the information of inter-relationship among different profile channels. There is little research in the literature on analyzing multi-channel profiles. In this paper, for the purpose of monitoring and fault diagnosis, we propose a method for analyzing multi-channel profile based on uncorrelated multilinear principal component analysis. We show the effectiveness of the proposed method by using a case study on a multi-operation forging process.

This paper focuses the problem of modeling manufactured surfaces for statistical process control. The application of Multilinear principal component analysis (MPCA) is introduced. MPCA is the generalization of the regular principal component analysis where the input can be not only vectors, but also tensors. The objective of this work is basically to explore the MPCA, as well as some basic concepts of multilinear algebra, for modeling manufactured surfaces. A real case study concerning cylindrical surfaces obtained by a lathe-turning process is taken as reference. The measurements related to a specific surface are stored in a matrix addressed by 2 index variables, while the observed data set related to several surfaces is stored in a 3rd- order tensor addressed by 3 indexes. Since the targeted application involves only the use of 3rd-order tensors of real entries, in this study the implementation of MPCA is limited to this specific case. Although a specific geometry is used herein as reference case study, any 2.5- dimensional surface (i.e. where scalar measurements are sampled for each item by using a fixed grid of two spatial index variables) can be modeled with the proposed MPCA-based approach.

An increasing amount of commercial measurement instruments implementing a wide range of measurement technologies is rapidly becoming available for dimensional and geometric verification. Multiple solutions are often acquired within the shop-floor with the aim of providing alternatives to cover a wider array of measurement needs, thus overcoming the limitations of individual instruments and technologies. In such scenarios, multisensor data fusion aims at going one step further by seeking original and different ways to analyze and combine multiple measurement datasets taken from the same measurand, in order to produce synergistic effects and ultimately obtain overall better measurement results. In this work an original approach to multisensor data fusion is presented, based on the development of Gaussian process models (the technique also known as kriging), starting from point sets acquired from multiple instruments. The approach is illustrated and validated through the application to a simulated test case and two real-life industrial metrology scenarios involving structured light scanners and coordinate measurement machines. The results show that not only the proposed approach allows for obtaining final measurement results whose metrological quality transcends that of the original single-sensor datasets, but also it allows to better characterize metrological performance and potential sources of measurement error originated from within each individual sensor.

This paper explores advantages arising from properly combining information provided by two sensors (contact and non-contact one) when measuring the same feature. When both of the metrology devices are used in cooperation, datasets of different resolution (a.k.a. multi-resolution data) have to be properly integrated in order to reconstruct the measured surface (or any geometric feature of interest) in both the sampled and the unsampled locations. To this aim, we propose a two-stage model, which consists of a low-resolution data model and a linkage model connecting the low- and the high-resolution data. The low-resolution data model is a spatial statistics model, specifically a Gaussian Process (GP). The linkage model has been adapted from the literature on calibrating computer simulation models of different accuracies. The newly developed two-stage model is used for quality inspection, showing that a model that properly combines multisensor information produces better results in terms of form error assessment when compared to models based on each single-resolution dataset, or with both but without structuring an appropriate data fusion model.

In modern industry, the development of complex products involves engineering changes that frequently require redesigning or altering the products or their components. In an Engineering Change process, Engineering Change Requests (ECRs) are natural language written texts exchanged among process operators. ECRs describe the required change on a product or a component and the solution. After the change implementation, ECRs are stored but never consulted, missing opportunities to learn from previous projects. This paper explores the application of text clustering to natural language texts written during the Engineering Change process in industry. In detail, the use of Self Organizing Map (SOM) to the problem of unsupervised clustering of ECR texts is explored. A case study is presented in which ECRs collected during the Engineering Change process of a railways industry are analysed. The results show that SOM text clustering has a good potential to improve overall knowledge reuse and exploitation.

Advanced 3D metrology technologies such as Coordinate Measuring Machine (CMM) and laser 3D scanners have facilitated the collection of massive point cloud data, beneficial for process monitoring, control and optimization. However, due to their high dimensionality and structure complexity, modeling and analysis of point clouds is a challenge. In this paper, we utilize techniques developed in multilinear algebra and propose a set of tensor regression approaches to model the variational patterns of point clouds and link them to process variables. The performance of the proposed methods is evaluated through simulation and a real case study of turning process optimization.

In multisensor coordinate metrology scenarios involving the fusion of homogenous data, specifically 3D point clouds like those originated by CMMs and structured light scanners, the problem of registration, i.e. The proper localization of the clouds in the same coordinate system, is of central importance. For fine registration, known variants of the Iterative Closest Point (ICP) algorithm are commonly adopted; however, no attempt seems to be done to tweak such algorithms to better suit the distinctive multisensor nature of the data. This work investigates an original approach that targets issues which are specific to multisensor coordinate metrology scenarios, such as coexistence of point sets with different densities, different spatial arrangements (e.g. sparse CMM points vs. gridded sets from light scanners), and different noise levels associated to the point sets depending on the metrological performances of the sensors involved. The proposed approach is based on combining known ICP variants with novel point set augmentation techniques, where new points are added to existing sets with the purpose of improving registration performance and robustness to measurement error. In particular, augmentation techniques based on advanced fitting solutions promote a paradigm shift for registration, which is not seen as a geometric problem consisting in moving point sets as close as possible to each other, but as a problem where it is not the original points, but the underlying geometries that must be brought together. In this work, promising combinations of ICP and point augmentation techniques are investigated through the application to virtual scenarios involving synthetic geometries and simulated measurements. Guidelines for approaching registration problems in industrial scenarios involving multisensor data fusion are also provided.

Accurate cost prediction during a new product development process is an important factor influencing the ability of manufacturing firms to survive in today’s competitive markets. Various methods have been proposed to predict costs during new product development processes. Although most of these methods produce point estimates, in practice, it is more realistic and useful for a method to provide interval predictions. In this paper, the application of Ordinary Least Squares (OLS) regression is considered in order to model available past data of production cost, and thus to compute interval predictions as well as point estimates for similar new products. With reference to a real case study in manufacturing, the issue of model selection is considered and a comparison of some commonly used criteria (functions of model fit penalized for model complexity) for selecting the best regression equation is analyzed. The objective of this study is to discuss the effect that the criterion used for selecting the best regression equation could have on the precision and accuracy of the prediction. Labor time data concerned the development of new sofa models were exploited as reference test case. The OLS regression is applied to several sets of labor time data and validated with respect to its fitting and predictive accuracy. The results of this study give insight into the efficacy of OLS regression for production cost estimation, and evidence on which criterion appears suitable to be used for model selection.

Continuous advances of sensor technology and real-time computational capability are leading to data-rich environments to improve industrial automation and machine intelligence. When multiple signals are acquired from different sources (i.e. multi-channel signal data), two main issues must be faced: (i) the reduction of data dimensionality to make the overall signal analysis system efficient and actually applicable in industrial environments, and (ii) the fusion of all the sensor outputs to achieve a better comprehension of the process. In this frame, multi-way principal component analysis (PCA) represents a multivariate technique to perform both the tasks. The paper investigates two main multi-way extensions of the traditional PCA to deal with multi-channel signals, one based on unfolding the original data-set, and one based on multi-linear analysis of data in their tensorial form. The approaches proposed for data modelling are combined with appropriate control charting to achieve multi-channel profile data monitoring. The developed methodologies are demonstrated with both simulated and real data. The real data come from an industrial sensor fusion application in waterjet cutting, where different signals are monitored to detect faults affecting the most critical machine components.

While in the previous chapters different approaches for quality monitoring of geometric tolerances were discussed, in the present one a comparison study is provided in which all of these approaches are considered. The aim is to allow practitioners to select a specific method in a given situation. A manufacturing reference case study is first detailed, namely, profiles measured on machined items subject to geometric specification (roundness). Then, two simulation scenarios are considered for the comparison study, where each scenario is obtained by perturbing the real case study. Competing approaches are ranked in each production scenario considering as a performance index the quickness in detecting out-of-control shapes.

Chapter 8 In the SPC-related literature, there are various cases in which we are interested in monitoring geometric specifications of products. Roundness, flatness, straightness, or cylindricity of products could be examples of such quality characteristics. It is obvious that under such circumstances, one would be interested in detecting any departure from the baseline shape of the product. Figure 1.17 shows various types of nonconforming shapes for a machined cylinder discussed by Henke et al. (1999) and Zhang et al. (2005). Detailed discussions on the methods for monitoring quality characteristics related to the geometric specifications of manufactured products are presented in this chapter.

We face the problem of monitoring complex shapes machined by a manufacturing pro- cess. The proposed method is based on combining a statistical model of the manufactured surface with multivariate control charting. In particular, a Gaussian Process model of the in-control surface pattern is firstly used as baseline. Then, multivariate control chart is used to monitor the difference between predicted and actual surface data. This procedure is compared with another method recently proposed in the literature. Two simulated case studies are used as reference test cases, throughout the paper.

This chapter shows how traditional approaches for statistical process control can be used for monitoring geometric tolerances. These approaches can be useful to quickly detect changes in the manufacturing process. In particular, two simple methods are presented. The first one uses standard variable control charts for monitoring over time the error associated with the geometric tolerance at hand. The second approach designs a control band around the mean shape of the feature associated with the geometric tolerance. Both approaches are shown with reference to the problem of monitoring a roundness form tolerance. Given their ease of use, the approaches are viable solutions for industrial practice.

Due to the easy accessibility of the 3D metrology tools such as Coordinate Measuring Machine or scanning tools, structured point cloud data is becoming more and more popular. Therefore, modeling the structured point cloud is an important task in many application domains. We model the structure point cloud as tensor and propose regularized tucker decomposition and regularized tensor regression to detect the variation patterns in the data and link these patterns to the process variables. Furthermore, the performance of the proposed method is evaluated through simulation and a real case study in the point cloud data in the turning process.

Le tecnologie che sfruttano metodi di rilevamento senza contatto sono sempre più utilizzate per migliorare l’efficienza dell’ispezione qualità in ambito manifatturiero. E’ evidente come, i recenti progressi dei sistemi di scansione laser e di visualizzazione hanno fatto si che sia possibile migliorare anche l’efficienza del controllo statistico dei processi di produzione. In questo articolo, gli autori presentano un sistema messo a punto per fornire in modo rapido misure di interesse delle features geometriche, che normalmente caratterizzano gli imbutiti da lamiere piane, integrato ad un modulo software di elaborazione delle misure per automatizzare l’ispezione della qualità e il controllo statistico di processo. Il sistema è costituito da componenti hardware e software appositamente sviluppati per le finalità di utilizzo della macchina messa a punto. La parte hardware è costituita da un’apparecchiatura a scansione laser e dall’attrezzatura che consente di riferire opportunamente i componenti al sistema durante la fase di misurazione. La parte software ha la finalità di generare part program di scansione ottimizzati in funzione della geometria del componente. Ciascuna scansione viene registrata e va a fare parte di un database che opportunamente organizzato consente di fornire informazioni utili per il controllo della qualità e il monitoraggio del processo di produzione preso in esame. Il sistema realizzato è stato progettato per consentire il rilevamento dei componenti a bordo linea. In questo articolo, gli autori, oltre a descrivere nello specifico le principali caratteristiche del sistema di acquisizione messo a punto, riportano i risultati del suo utilizzo su di un caso pratico dato dall’analisi di un lotto di produzione di un componente automobilistico ottenuto per imbutitura di lamiera piana.