Design optimization method of pipeline parameter based on improved artificial neural network

The rationality of pipeline design is directly related to its energy efficiency, reliability, and safety. Pipeline vibration may lead to negative effects such as mechanical loss and fatigue damage. Therefore, this study utilizes pipeline optimization design to mitigate these effects. Recently, neural networks have been widely used in structure design optimization. In the study, a backpropagation neural network (BP) combined with a variant slime mould algorithm (SMA) is utilized to solve the pipeline structure design optimization problem. Pipeline transport plays a crucial role in the efficient movement of various commodities, including but not limited to gas, oil, water, and other liquid substances. The interaction between liquid and pipeline can cause pipeline vibration and even damage. Therefore, based on the simulation model considering FSI (fluid-structure interaction), machine learning methods such as BP can predict vibration characteristics of fluid-conveying pipelines. However, existing research has shown that BP has insufficient parsing ability in structure mechanics problems, especially in solving the overall characteristics of complex structures (such as maximum structural strain). This study proposes an Arithmetic-based slime mould algorithm (ACSMA) with an adaptive decision strategy and a chaotic mapping strategy. A hybrid algorithm named ACSMA-BP is presented to promote the model's prediction ability. At last, to verify the effectiveness of the proposed pipeline structure design optimization approach, the ACSMA-BP model is utilized to complete a structure design optimization case for a simulated pipeline. The numerical results indicate that compared with AOA, CWOA, ESSAWOA, NGS_WOA, and RSA, the ACSMA has the best optimization ability.