Progress 09/15/03 to 09/14/07

Outputs
OUTPUTS: The optimization portion of this project, related to moist-air convection systems for commercial cooking of meat products, consisted of: (1) Evaluating various strategies for determining optimal process profiles, and (2) Applying those techniques to commercially-relevant case studies. Three process models were used: a finite element model of the process and two neural network models (static and dynamic) developed by using data produced by the finite element model. Three global optimization algorithms (genetic algorithm, simulated annealing, and integrated controlled random search - ICRS) were combined with those models. Discrete (piecewise linear interpolation) and continuous (Fourier series) control parameterizations were tested. Combined strategies were tested to find optimal dynamic control profiles (temperature, humidity, impingement velocity, and cooking duration) maximizing patty yield while satisfying safety (Salmonella inactivation) and quality (internal and surface color) constraints for ground beef patties (20% fat). In addition, sensitivity of the optimal solutions was studied. In applying the optimization results to commercial case studies, the finite element model was coupled with the three global optimization algorithms to optimize single-stage, double-stage, and multi-zone impingement oven systems. The multi-zone oven was a hypothetical, modified single-stage system, in which the air velocity could be controlled in four distinct zones within the oven. Because of imperfect knowledge about uncertainty in component models, particularly those for color development, subsequent projects have been initiated to resolve this limitation. Initial tests were conducted on location at a manufacturer of commercial oven systems, in a commercial-scale, moist-air oven system, to collect data on product temperature, surface moisture content, and surface color (via a Hunter colorimeter). Those data were used to parameterize a 0th-order kinetic model, with a modified, Arrhenius-type model for the effect of temperature and moisture on color. A subsequent proposal, focused only on this research question, has already been written and submitted, in order to refine the applicability of the completed optimization protocols to commercial systems. With respect to the graduate education component of the project, the original web-based graduate course, focused on microbial modeling (at MSU), was significantly modified to remove overlap with the new courses, and piloted in the new form in 2006, including new information about modeling techniques. The new course on non-thermal processing technology (at OSU) was refined/completed as part of the transition from traditional to hybrid to fully-web-based. Plans for implementing the graduate certificate in Food Safety Engineering were delayed, due to the loss of a non-funded collaborator. However, a new collaborator was identified, and these plans will move forward beyond the duration of this funded project. PARTICIPANTS: Dr. Bradley Marks was the project director for this project and was responsible for managing project progress and supervising the MSU personnel (other than the co-PIs) who worked on this project. He also modified the web-based graduate course on microbial modeling and conducted the preliminary color development tests and modeling on-site at a manufacturer of commercial oven systems. Drs. Elliot Ryser and Alden Booren oversaw the microbial methodologies and meat sample/process evaluations, respectively. Dr. Sanghyup Jeong successfully completed a Ph.D. in biosystems engineering at MSU as part of this project; he developed and tested the optimization programming and models that were central to the research elements of the project. Dr. Jeong is continuing as a visiting assistant professor at MSU, where is writing the manuscripts reporting the final results of this project and extending the work to new, related projects. Dr. Alicia Orta-Ramirez and Nicole Hall were a research assistant professor and a research specialist who provided support in the experimental aspects of the project. Dr. Danilo Campos and Biao Guo, a Ph.D. student in food science, both worked on development of a protocol for an industry survey to evaluate general practice for thermal process validation. Lastly, three undergraduate students worked on various aspects of this project, in support of the senior personnel and graduate students. At Ohio State University, Dr. V.M. (Bala) Balasubramaniam and his post-doctoral associate developed new, web-based instructional resources covering non-thermal technologies related to engineering food safety. TARGET AUDIENCES: There are three target audiences for this project. Manufacturers of commercial oven systems will benefit from tools to improve oven design. Meat and poultry processors will benefit from improved tools for ensuring product safety and improving processing yield. Lastly, practicing engineers in the industry will benefit from greater access to formal instructional resources supporting food safety obligations in this sector.

Impacts
Overall, the results of this project showed that specific optimization strategies could be applied to complex, multi-stage oven cooking systems. However, the optimization strategies based on the two neural network models showed potential to commit false-pass classification errors (7% and 15%), which can result in dangerous process solutions. Also, the optimal control profiles were found to be very sensitive (100% failure) to reasonable perturbations. However, sub-optimal solutions were less sensitive (85% and 49% failure) to the same perturbations. The optimal dynamic process profiles for maximizing yield suggested that 2- or 3-step processes might approximate the ideal control profiles. These comparisons and pitfalls for dynamic process optimization provide an important foundation to develop more effective and practical techniques for complex food processing operations. In the specific test cases, the double-stage oven was identified as the most effective system, achieving 67% cooking yield for ground beef patties, with a relatively fast process time (half that of the single-stage oven). The multi-zone oven was the second best (65% yield). Even though the numerical results are highly dependent on factors such as product properties, the optimization methods were effective in finding optimal solutions satisfying actual industrial demands. The results also suggest potential changes in oven design or operation, which might be beneficial to oven manufacturers and processors. The preliminary work on color modeling, which is an extension of the original project proposal, has indicated that surface moisture content is a critical factor in predictive modeling of surface browning, and that such modeling is feasible, given sufficient future data collection. Overall, it is anticipated that the computational tools and training resources resulting from this project will enable meat and poultry processors to increase their certainty in meeting performance standards for microbial safety of ready-to-eat products, while simultaneously increasing cooking yield and thereby economic returns. Additionally, the graduate certificate program that is being developed ultimately will enable practicing engineers in the food industry to enhance their knowledge base and credentials in the area of food safety.

Publications

• No publications reported this period

Progress 01/01/06 to 12/31/06

Outputs
The results of 2005 yielded significant progress in application of a process model and various optimization strategies for moist-air convection systems for commercial cooking of meat products. The optimization methods were effective in finding optimal solutions; however, those solutions were vulnerable to process failure, because of imperfect knowledge about variability and uncertainty in component models, particularly those for color development. Therefore, in 2006, plans have been made to extend the work of this existing project, in order to enable the optimization methods to yield solutions that are of industrial relevance. In particular, quality prediction models will be refined, and the statistical confidence of the component and integrated process models will be quantified. With respect to the graduate education component of the project, the original web-based graduate course, focused on microbial modeling (at MSU), was significantly modified to remove overlap with the new courses, and piloted in the new form in 2006. The new course on non-thermal processing technology (at OSU) is continuing to be refined in transition from traditional to hybrid to fully-web-based, as plans for initiating the web-based graduate certificate program take shape in 2007.

Impacts
It is anticipated that the computational tools and training resources resulting from this project will enable meat and poultry processors to increase their certainty in meeting performance standards for microbial safety of ready-to-eat products, while simultaneously increasing cooking yield and thereby economic returns. Additionally, the graduate certificate program will enable practicing engineers in the food industry to enhance their knowledge base and credentials in the area of food safety.

Publications

• Jeong, S. 2005. Process optimization for moist air impingement cooking of meat patties. Ph.D. dissertation. Michigan State University. East Lansing, MI.
• Jeong, S., Marks, B.P.. 2006. Application of process optimization techniques to cooking of meat patties in industrial, moist-air impingement ovens. IFT Abstract 078D-11. Presented at the Institute of Food Technologists Annual Meeting. Orlando, FL. June, 2006.
• Jeong, S., Marks, B.P. 2006. Evaluation of strategies using neural networks, global optimization algorithms, and control parameterization for multivariable dynamic process optimization. IFT Abstract 078E-07. Presented at the Institute of Food Technologists Annual Meeting. Orlando, FL. June, 2006.

Progress 01/01/05 to 12/31/05

Outputs
Significant progress has been made in two areas of this project, related to optimizing moist-air convection systems for commercial cooking of meat products: (1) Evaluating various strategies for determining optimal process profiles, and (2) Applying those techniques to commercially-relevant case studies. Three process models were used: a finite element model of the process and two neural network models (static and dynamic) developed by using data produced by the finite element model. Three global optimization algorithms (genetic algorithm, simulated annealing, and integrated controlled random search - ICRS) were combined with those models. Discrete (piecewise linear interpolation) and continuous (Fourier series) control parameterizations were tested. Combined strategies were tested to find optimal dynamic control profiles (temperature, humidity, impingement velocity, and cooking duration) maximizing patty yield while satisfying safety (Salmonella inactivation) and quality (internal and surface color) constraints for ground beef patties (20% fat). In addition, sensitivity of the optimal solutions was studied. The optimization strategies based on the two neural network models showed potential to commit false-pass classification errors (7% and 15%), which can result in dangerous process solutions. Also, the optimal control profiles were found to be very sensitive (100% failure) to reasonable perturbations. However, sub-optimal solutions were less sensitive (85% and 49% failure) to the same perturbations. The optimal dynamic process profiles for maximizing yield suggested that 2- or 3-step processes might approximate the ideal control profiles. These comparisons and pitfalls for dynamic process optimization provide an important foundation to develop more effective and practical techniques for complex food processing operations. In applying these results to commercial case studies, the finite element model was coupled with the three global optimization algorithms to optimize single-stage, double-stage, and multi-zone impingement oven systems. The multi-zone oven was a hypothetical, modified single-stage system, in which the air velocity could be controlled in four distinct zones within the oven. The double-stage oven was identified as the most effective system, achieving 67% cooking yield with a relatively fast process time (half that of the single-stage oven). The multi-zone oven was the second best (65% yield). Even though the numerical results are highly dependent on factors such as product properties, the optimization methods were effective in finding optimal solutions satisfying actual industrial demands. The results also suggest potential changes in oven design or operation, which might be beneficial to oven manufacturers and processors. With respect to the graduate education component of this project, a third collaborator has been found to contribute the sanitation portion of the web-based graduate certificate program. The marketing/management plan for this program in Food Safety Engineering will be developed in 2006, in preparation for the first group of participants in 2006-2007.

Impacts
It is anticipated that the computational tools and training resources resulting from this project will enable meat and poultry processors to increase their certainty in meeting performance standards for microbial safety of ready-to-eat products, while simultaneously increasing cooking yield and thereby economic returns. Additionally, the graduate certificate program will enable practicing engineers in the food industry to enhance their knowledge base and credentials in the area of food safety.

Publications

• No publications reported this period

Progress 01/01/04 to 12/31/04

Outputs
To find optimal operating conditions for moist-air impingement cooking of low fat ground beef patties, a generalized regression neural network (GRNN) was utilized as an alternative process model. The GRNN was trained with training data sets generated by a pre-existing finite element model (FEM) and could predict final patty states (yield and lethality) with moderate accuracies. The GRNN was coupled with a genetic algorithm (GA) to locate sub-optimal conditions maximizing product yield while ensuring microbial safety. The neural network coupled GA optimization technique was applied to single stage, double stage, and multi-zone controlled ovens. The combined optimization method found the sub-optimal conditions for each process type. For a single stage oven, 83% yield was found as a sub-optimal solution. However, these preliminary optimization results did not account for product quality factors, such as color or texture, and therefore the initial optimization results suggested somewhat unrealistic solutions (e.g., operating with pure steam in the convection oven). Currently, the optimization techniques are being refined, and quality factors are being included as constraints in the optimization programming, in order to generate results that are directly relevant to the ready-to-meat industry. With respect to the graduate education component of this project, a new, completely web-based graduate course is under development in the area of non-thermal processing, which will be first offered in 2005. The marketing plan for the graduate certificate program in "Food Safety Engineering" will also be developed in 2005, in preparation for the first group of participants in 2005-2006.

Impacts
It is anticipated that the computational tools and training resources resulting from this project will enable meat and poultry processors to increase their certainty in meeting performance standards for microbial safety of ready-to-eat products, while simultaneously increasing cooking yield and thereby economic returns. Additionally, the graduate certificate program will enable practicing engineers in the food industry to enhance their knowledge base and credentials in the area of food safety.

Publications

• Jeong, S., Marks, B.P. 2004. Optimizing multi-stage air impingement cooking of meat patties. Paper 04-3022. American Society of Agricultural Engineers. St. Joseph, MI.

Progress 01/01/03 to 12/31/03

Outputs
This is a new project (09/15/2003). A graduate research assistant and post-doctoral research associate are in place, and research activities have been initiated. In particular, the optimization methods to be used are currently being tested with simplified cooking models, before applying them to the actual multi-stage process models to be used in the project.

Impacts
This is a new project (09/15/2003). It is anticipated that the tools resulting from this project will enable meat and poultry processors to increase cooking yield of ready-to-eat products while simultaneously increasing certainty in meeting performance standards for microbial safety of the products.

Publications

• No publications reported this period