Honglu Yu

Data-based latent variable methods for process analysis, monitoring and control

This paper gives an overview of methods for utilizing large process data matrices. These data matrices are almost always of less than full statistical rank, and therefore, latent variable methods are shown to be well suited to obtain useful subspace models from them for treating a variety of important industrial problems. An overview of the important concepts behind latent variable models is presented and the methods are illustrated with industrial examples in the following areas: (i) the analysis of historical databases and trouble-shooting process problems; (ii) process monitoring and FDI; (iii) extraction of information from novel multivariate sensors; (iv) process control in reduced dimensional subspaces. In each of these problems, latent variable models provide the framework on which solutions are based.

Multivariate image analysis and regression for prediction of coating content and distribution in the production of snack foods

Abstract An important problem in the snack food industry is to control the amount of coating applied to the base food product and the distribution of the coating among the individual product pieces. Multivariate image analysis and regression approaches based on Principal Component Analysis (PCA) and Partial Least Squares (PLS) are presented for the extraction of features from RGB (red–green–blue) color images and for their use in predicting the average coating concentration and the coating distribution. Data collected using both on-line and off-line imaging from several different snack food product lines are used to develop and evaluate the approaches. The methods are now being used in the snack food industry for the on-line monitoring and feedback control of product quality. This paper reports on the development of the methods.

Multivariate image analysis for process monitoring and control

Information from on-line imaging sensors has great potential for the monitoring and control of quality in spatially distributed systems. The major difficulty lies in the efficient extraction of information from the images, information such as the frequencies of occurrence of specific and often subtle features, and their locations in the product or process space. This paper presents an overview of multivariate image analysis methods based on Principal Component Analysis and Partial Least Squares for decomposing the highly correlated data present in multi-spectral images. The frequencies of occurrence of certain features in the image, regardless of their spatial locations, can be easily monitored in the space of the principal components. The spatial locations of these features can then be obtained by transposing highlighted pixels from the PC score space into the original image space. In this manner it is possible to easily detect and locate even very subtle features from on-line imaging sensors for the purpose of statistical process control or feedback control of spatial processes. The concepts and potential of the approach are illustrated using a sequence of LANDSAT satellite multispectral images, depicting a pass over a certain region of the earths surface. Potential applications in industrial process monitoring using these methods will be discussed from a variety of areas such as pulp and paper sheet products, lumber and polymer films.

Digital Imaging for Process Monitoring and Control with Industrial Applications

Abstract The development of on-line digital imaging systems for process monitoring and control is illustrated through two industrial applications: i) the control of coating concentration and distribution on snack food products. and ii) the monitoring of boiler systems through imaging of the combustion processes. Feature information extracted from images using Multivariate Image Analysis (MIA) based on Principal Component Analysis (PCA). is used to develop models to predict product quality and process property variables. The imaging systems are used to monitor these product quality and process property variables, to detect and diagnose operational problems in the plants, and to directly implement closed-loop feedback control.

Online prediction of organileptic data for snack food using color images

In this paper, a study for the prediction of organileptic properties of snack food in real-time using RGB color images is presented. The so-called organileptic properties, which are properties based on texture, taste and sight, are generally measured either by human sensory response or by mechanical devices. Neither of these two methods can be used for on-line feedback control in high-speed production. In this situation, a vision-based soft sensor is very attractive. By taking images of the products, the samples remain untouched and the product properties can be predicted in real time from image data. Four types of organileptic properties are considered in this study: blister level, toast points, taste and peak break force. Wavelet transform are applied on the color images and the averaged absolute value for each filtered image is used as texture feature variable. In order to handle the high correlation among the feature variables, Partial Least Squares (PLS) is used to regress the extracted feature variables against the four response variables.

Multivariate Image Analysis for Inferential Sensing: A Framework

Abstract This paper presents a framework for developing vision-based inferential sensors. This framework not only gives a summary of existing methodologies, but also combines the methods used in other areas, such as traditional machine vision, multivariate image analysis and multivariate data analysis, and gives a broad vision for future developments of the area.