Optimal animal nutrition

Linear programming has long been used in pig production to generate minimum cost diets that are subject to nutritional constraints. More recently, the development of accurate animal growth models and the development of effective nonlinear optimisation techniques, in conjunction with linear programming, have allowed us to address the bigger problem of building feeding schedules that maximise profitability. These feeding schedules, which we name "optimal feeding schedules", simultaneously minimise feed costs while maximising gross return. -- The random search for the optimal feeding schedule runs over an extremely large space. The optimisation methods used to generate the optimal solution are of critical importance. Software to perform the optimization, named "Bacon Max", has been developed in Visual C++. The search processes used in Bacon Max to determine the optimum solution for this problem include pure random search, tabu search, Monte Carlo search, ascent search and a genetic algorithm. Among these search processes genetic algorithms were the most successful, despite the long search period needed. This thesis extends this work in four ways. -- First (in Chapter 3), an alternative optimisation methodology was developed, which we call the "tailored" method, adapted to the known nature of the objective function, found by examining several random cross sections through a known solution. This revealed a single craggy peak. The tailored method finds the optimum more rapidly than the genetic algorithm. The program for the tailored method is also developed in Visual C++. -- Second (in Chapter 4), profitability depends on feed costs and price at slaughter which in turn are subject to variation across time. We show how to include such variation in our modelling and handle it using stochastic programming. The optimal feeding schedule is influenced by pig type, costs and prices, and dietary restraints. The effects are discussed. The Bacon Max software is developed to handle variations of feed cost and price at slaughter using stochastic programming. -- Third (in Chapter 5), the methodology is utilized to explore the impact of pig type, costs and dietary restraints on the optimal feeding schedule. At the same time the optimal feeding schedule is compared to the commonly used "feed-to-lean" schedule. -- Fourth (in Chapter 6), an application to real data from southern Thailand is presented. The results indicate that profitability can be increased by changing the original feeding schedule to the optimal feeding schedule resulting from Bacon Max. -- In summary, results indicate that the optimal feeding schedule from Bacon Max can improve pro tability for the producer.