GROWING TREE METHOD FOR OPTIMISATION OF MULTIFACTORIAL EXPERIMENTS

Context. The task of planning multifactorial experiments is important in science and industrial production. In the context of competition, rising costs, and increasing efficiency, it is necessary to optimize plans for multifactorial experiments in terms of cost and time. To solve this problem, there are a number of approaches and methods, the choice of which for a competitive technical task is an important and difficult task. In this regard, there is a need to develop new methods for optimizing the cost (time) of multifactorial experiment plans, compare them with existing methods, and give recommendations for practical application in the study of real objects. Objective. The purpose of the study is to develop and test the method of growing trees, to evaluate its effectiveness in comparison with other methods. The following tasks has been solved to achieve this goal: the proposed method of growing trees has been implemented in the form of software; the method has been used to optimize plans for multifactorial experiments in the study of real objects; its effectiveness has been evaluated in comparison with other methods; recommendations for its use were given. Method. The proposed method of growing trees is based on the application of graph theory. The advantage of the method is the reduction of time for solving optimization problems related to the construction of optimal plans for multifactorial experiments in terms of cost (time) expenses. Another characteristic feature is the high accuracy of solving optimization problems. Results. The results of experiments and comparisons with other optimization methods confirm the efficiency and effectiveness of the proposed method and allow us to recommend it for the study of objects with the number of significant factors k ≤ 7. It is promising to further expand the range of scientific and industrial objects for their study using this method. Conclusions. A growing tree method has been developed for the optimization of multifactorial experimental plans in terms of cost and time expenditures, along with software that implements it using the Angular framework and the TypeScript programming language. The effectiveness of the growing tree method is shown in comparison with the following methods: complete and limited enumeration, monkey search, modified Gray code application, and bacterial optimization. The growing tree method is faster than complete enumeration and can be applied to optimize multifactorial experimental plans in terms of cost (time) expenses for objects with a number of factors k ≤ 7. In solving optimization problems, the method of growing trees gives better results compared to monkey search, limited enumeration and bacterial optimization.