Status of Observational Models Used in Design and Control of Products and Processes Shyam S. Sablani This article is part of a collection entitled “Models for Safety, Quality, and Competitiveness of the Food Processing Sector,” published in Comprehensive Reviews in Food Science and Food Safety. It has been peer-reviewed and was written as a follow-up of a pre-IFT workshop, partially funded by the USDA NRI grant 2005-35503-16208. ABSTRACT: Modeling techniques can play a vital role in developing and characterizing food products and processes. Physical, chemical, and biological changes that take place during food and bioproduct processing are very complex and experimental investigation may not always be possible due to time, cost, effort, and skills needed. In some cases even experiments are not feasible to conduct. Often it is difficult to visualize the complex behavior of a data set. In addition, modeling is a must for process design, optimization, and control. With the rapid development of computer technology over the past few years, more and more food scientists have begun to use computer-aided modeling techniques. Observation-based modeling methods can be very useful where time and resources do not allow complete physics-based understanding of the process. This review discusses the state of selected observation- based modeling techniques in the context of industrial food processing. Introduction Observational models are referred to as data-driven models. They are primarily empirical in nature and are inferred primarily from measured data. Observational models are “black box” mod- els to different degrees in relation to the physics of the process. The classical statistical models can have a model in mind (often based on some understanding of the process) before obtaining the measured data. This makes them less of a black box than models such as neural network or genetic algorithm (GA) that are frequently completely data driven, no prior assumption is made about the model, and no attempt is made to physically interpret the model parameters once the model is built (Datta and Sablani 2006). There are many practical situations where time and resources do not permit complete physics-based understanding of a pro- cess. Physics-based models often require more specialized train- ing and/or longer development time. In some applications, de- MS 2007-0243 Submitted 4/5/2007, Accepted 9/4/2007 . The author is with the Dept. of Biological Systems Engineering, Washington State Univ., PO Box 646120, Pullman, WA 99164–6120, U.S.A. Direct inquiries to author Sablani (E-mail: ssablanis@wsu.edu ). tailed understanding provided by the physics-based model may not even be necessary. For example, in process control, detailed physics-based models are often not needed and observation- based models can suffice. Observation-based models can espe- cially be useful when relationships between process parameters and quality/safety are not clear. These models can be extremely powerful in providing practically useful relationships between input and output parameters for complex processes. The types of data available and the purpose of modeling usu- ally influence the kind of observation models to be used (Table 1). Several observational models have been developed over the years (Figure 1). Statistical models for design of experiments (DE) orig- inated in the 1920s to maximize the knowledge gained from the experimental data and it has evolved over the last 75 y. Response surface techniques associated with design of experimental meth- ods for the purpose of optimization were introduced in the 1950s. Both statistical techniques have been used extensively in the food literature for the design of processes and products. The advent of high-speed computers with large storage facility and development of easy-to-use software have led to an implementation of the more sophisticated statistical technique of multivariate analysis (MVA). Over the last decade, MVA techniques have seen recognition in the area of characterization of food quality, and this is due to 130 COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETY—Vol. 7, 2008 C  2008 Institute of Food Technologists