基于混合神经网络的门座起重机变幅机构参数优化设计

门座起重机变幅机构优化设计通常采用基于实例推理的方法确定优化设计初始参数,然后再对初始参数进行优化,得出最后的优化设计参数。这种基于实例优化方法的主要问题在于,难于确定相似实例和难于将相似实例应用于当前的实例中,并且由此确定的优化初始参数只偏向于某一个特定的实例,不具备某类实例的普遍特性。提出基于混合神经网络变幅机构优化设计方法,该方法采用一种混合神经网络,可用来确定优化设计的初始参数。这种方法计算上更为简单直观,对于训练好的混合神经网络,可直接由输入参数得到设计初始参数。这种初始参数并带有某类实例的一般特性,对其进行优化可得到较好的优化结果。

HYBRID NEURAL NETWORKS BASED PORTAL CRANES' LUFFING SYSTEM OPTIMAL DESIGN

Traditional portal cranes' luffing system design process generally includes employing case-based reasoning method for reasonable initial parameters, and then optimizing them to get the optimal results. But the approach is not desirable because it's hard to decide which case is the nearest and how to map the nearest case to the current problem, also the initial parameters thereby are partial to a special case, without general attributes of some type of cases. A hybrid neural networks is presented based on optimization method for portal cranes' luffing system design, which is simple in computation, and by which the initial parameters obtained can lead to more desirable optimization results.