Abstract: Characteristic signal of solid propellant combustion more and more becomes an important factor to restrict missile stealth characteristics and guidance precision development. In order to solve the problem of long formulation design cycle to reduce the characteristic signal, which is caused by traditional experimental method, Particle Swarm Optimization Algorithm（PSOA） is studied to find the optimal design scheme of solid propellant formulations to reduce characteristic signal. In the process, the rejection method, the penalty function method and the strategy of keeping population diversity are used to improve standard PSOA properly. It solves the nonlinear constraint problem, overcomes the defects of algorithm to fall into local optimum easily and improves global search ability. Establishment of the formulation optimization model and simulation results show that in the aspect of reducing characteristic signal, improved PSOA is superior to some other intelligent algorithms, such as, improved genetic algorithm, standard particle swarm optimization and so on， and it can also shorten the formulation design cycle.

Key words: Particle Swarm Optimization Algorithm（PSOA）, reject method, penalty function method, strategy of keeping population diversity, formulation optimization model

摘要： 固体推进剂燃烧产生的特征信号越来越成为制约导弹武器隐身特性和制导精度发展的重要因素，为解决传统实验法降低特征信号而进行配方设计周期长的问题，采用粒子群算法寻找固体推进剂配方最优设计方案以达到降低特征信号的目的，同时综合运用拒绝法、罚函数法以及种群多样性保持策略对标准粒子群算法进行适当改进，解决了非线性约束问题，克服了算法容易陷入局部最优的缺陷，提高了全局搜索能力。建立配方优化数学模型并进行仿真，结果表明采用改进粒子群算法来降低固体推进剂特征信号，优于标准粒子群算法、改进遗传算法等智能算法，并能缩短配方设计周期。

关键词: 粒子群算法, 拒绝法, 罚函数法, 种群多样性保持策略, 配方优化数学模型