Investigation of the aerodynamic optimization design of fluid machinery based on machine learning

Fluid machinery plays an important role in national pillar industries such as national defense, military, aerospace, heavy industry, energy and power and is also the main industrial energy source. At present, many problems remain in the design of fluid machinery and systems. From the point of view of system optimization design, the utilization rate of fluid mechanics is low, which is mainly due to the mismatch between the system and the network. Based on this, relevant scholars have proposed changing high-pressure fluid machinery hydraulic systems into medium-pressure or low-pressure impeller systems and using hydraulic coupling to regulate the flow of mechanical pumps and fluid machinery to further achieve fluid machinery purposes. However, with the high efficiency, high precision and scalability of fluid machinery, a traditional single core processor has been unable to meet the high precision requirements. Therefore, how to develop software suitable for high-performance computing according to different actual conditions is an important problem. In recent years, with the development of machine learning technology, various algorithms have emerged and been widely used in fluid mechanics. These include the naive Bayesian classifier algorithm, K-means clustering algorithm, K-means clustering machine learning algorithm and support vector machine learning algorithm. Machine learning algorithms based on deep learning have a strong inductive learning ability. They can find the potential flow field information through a large number of experiments and numerical simulations. Through machine learning, a mathematical hydrodynamics gas optimization model is established. The research results show that on this basis, the overall pressure of the hydrodynamic gas optimization model established using a machine learning method was 10.1% higher than that of the control group. This showed that the hydrodynamic gas optimization model established using a machine learning method had better aerodynamic characteristics, providing a reference for future related work.