(26c) Auto Plot Pathfinder : Can Multi-Purpose Optimization Generate a Plant Layout Comparable to a Veteran Engineer ?

JGC has been developing a generative design method for plant layout -what we call Auto Plot PATHFINDER-, applying satisficing trade-off method implemented with genetic algorithm and JGC expert knowledge. It instantly generates numerous numbers of layout designs that either an engineer or an algorithm alone cannot generate.

The exploration in the generative designs is formulated as an area-constrained knapsack problem, in which process units are allocated to strike the trade-off balance between minimizing a piping cost and minimizing a layout area. The method is developed analogically based on the JGC experts’ way of plant layout design. It consists of a rule portion such as complying with maintainability, safety, and regulations, and an optimization portion such as pursuing economic efficiency and competitive advantage. The rules are developed with a design structure matrix that formulates expert knowledge, while the genetic algorithm works for the optimization portion. Setting the aspiration level of piping cost and layout area enables users like piping engineers to control the exploration space. Furthermore, the objective function has such flexibility that the users can embed their domain knowledge into the objective function. As characterized above, the practical generative design is realized by a hybrid optimization method of mathematical techniques and engineering expert knowledge.

A proof-of-concept trial where Auto Plot PATHFINDER has been demonstrated with the past design projects has experimentally proven it broadly applicable to plant design. We have confirmed users can manipulate the exploration space and the optimization result by setting the aspiration level. A Pareto-frontier of the 2-dimension case that comprises piping costs and layout area shows the plant layout generated and the asymptotic behavior under minimum piping cost and minimum layout area are found consistent with our experts’ intuition and insight, suggesting that the expert knowledge is successfully embedded into the algorithm. In addition, it surprisingly reproduces the creativity of veteran engineers. The analysis of the objective function shows us the very small techniques that veteran engineers often utilize make such a huge contribution to reducing piping costs as approximately 40% reduced. Auto Plot PATHFINDER can help us with discovering what makes a great plant layout great.