基于深度学习的传感器优化布置方法

针对传感器优化布置(optimal sensor placement,简称OSP)问题,提出了一种新的使用深度神经网络的解决方案,并以简化的桥梁形状的桁架结构中的振动测试传感器优化为例进行了验证。首先,选择一种传统的传感器优化布置方法,对自动化生成的大量不同的桁架结构分别进行传感器优化布置计算,将所得优化布置结果在进行数据预处理后构建出深度学习方法所需要的训练集与验证集;其次,使用Python语言和深度学习框架TensorFlow设计实现与本研究问题适配的深度神经网络模型并训练;然后,随机生成了新的桁架结构参数;最后,将深度神经网络输出的传感器布置结果和传统方法的计算结果进行了比较,验证了本研究方法的有效性以及在速度上、可移植性与可扩展性方面的性能优势。

Optimal Sensor Placement Based on Deep Learning Method

Aiming at the problem of optimal sensor placement (OSP), a new method based on the deep neural network is proposed, and the optimization of vibration test sensors in a simplified bridge-shaped truss structure is used for verification. First, a traditional sensor optimization arrangement method is employed to find the optimal sensor placement of a large number of automatically generated truss structures. The training set and verification set required by the deep learning method are constructed from the preprocessed optimization results. Second, a deep neural network model adapted to the problem studied in this paper is designed and trained using Python and the deep learning framework TensorFlow. Then, new truss structure parameters are randomly generated. Finally, the optimization results given by the deep neural network and the traditional method are compared, which verifies the feasibility and advantages in speed, portability and scalability of the sensor optimization layout scheme based on the deep learning method.