Abstract: To get the better solutions of the single objective engineering optimization problems, which have multi-parameters and multi-constraints, a novel particle swarm optimization algorithm is proposed with classification learning. The particle swarm is divided into three classes, i.e. better class, middle class and worse class. For each class, different learning methods and directions are used, respectively. For the better class, the learning speed and direction itself are remained. For the middle class, the interactive learning strategy is introduced. For the worse class, the learning direction to the better class is modified. Hence, the algorithm is not affected by the continuous and differentiable functions. It is illustrated that, by the numerical experiments, this algorithm has the better performance, to deal with the function which contains uni-modal, multi-modal discrete and dynamic problems, comparing with other improved particle swarm optimization algorithms. It is indicated by the engineering application examples that this algorithm can get the better parameters which can make the system to get the better performance, when dealing with structure design and hydraulic straightener PID controller problems.

Key words: classification learning, hydraulic straightener, particle swarm optimization, PID, structure design