基于BAS-BP神经网络的轮廓坐垫舒适性评价

Evaluation of contoured seat cushion comfort based on BAS-BP neural network

研究表明轮廓坐垫对提升舒适性有帮助,但不同轮廓形状坐垫对于舒适性的影响尚不明确。为更客观准确地研究不同轮廓坐垫形状与舒适性的关系,提出了一种基于天牛须算法(BAS)优化BP神经网络的轮廓坐垫舒适性评价模型。首先,开展了基于体压分布的轮廓坐垫舒适性主客观评价实验,获取客观实验数据和被试人员主观评价;其次,对实验数据进行预处理和统计计算,从而得到舒适性评价模型的实验数据集;再次,根据实验数据对轮廓坐垫舒适性影响因素进行简要分析,利用相关性分析确定神经网络的输入参数;最后,利用BAS优化BP神经网络,构建BAS-BP神经网络评价模型并实现对轮廓坐垫舒适性的评价。结果表明:BAS-BP神经网络计算结果比标准BP神经网络计算结果的均方误差降低了59.26%,决定系数提高了22%,BAS-BP神经网络的评价效果更为准确。该优化模型提高了舒适性评价的准确率,满足实际应用的需要,为轮廓坐垫的设计和产品开发早期阶段对不同形状坐垫舒适性的评价提供了理论研究基础及实际应用方法。

With the rapid development of science and technology and the gradual improvement of people’s living conditions, the importance of personal health has been deeply rooted in the hearts of the people. As the most common posture in contemporary life and work, sitting posture is closely related to human health, and users’ requirements for sitting comfort are constantly increasing. As a new product, contoured seat cushions can help improve comfort, but the effect of the different degree of contoured seat cushions on the comfort is still unclear. To study the relationship between different contoured seat cushion shapes and comfort more objectively and accurately, a contoured seat cushion comfort evaluation model based on the beetle antennae search(BAS) algorithm optimization of BP neural network was proposed. Firstly, the subjective and objective evaluation experiments of contoured seat cushion comfort based on body pressure distribution were carried out, and the objective experimental data and subjective evaluation of the subjects were obtained. Secondly, the experimental data was preprocessed using the statistical method and the comfort evaluation model was developed using the obtained experimental data. Thirdly, according to the experimental data, the factors affecting the comfort of the contoured seat cushion were briefly analyzed and the correlation analysis was used to determine the input parameters of the neural network. Finally, the BP neural network was optimized using the beetle antennae search algorithm, and the BAS-BP model was constructed to evaluate the comfort of the contoured seat cu-shion. Compared with the standard BP neural network, the mean square error of BAS-BP neural network was reduced by 59.26%, and the coefficient of determination was increased by 22%. The evaluation accuracy was significantly improved by the BAS-BP neural network. The optimization model improved the accuracy of comfort evaluation, met the needs of practical applications, and provided the theoretical research basis and practical application method for the design of contoured cushions and the evaluation of the comfort of different shapes of cushions in the early stages of pro-duct development. The study indicated that the shape of the contoured seat cushion would affect the pressure distribution and subjective comfort on the seat surface, so it is recommended that the design of the contoured seat cushion should fit the user’s thighs and buttocks as much as possible to improve the user’s comfort experience.