Using Surrogate Models in the Construction of a Pareto-Optimal Positioning Electropneumatic Actuator with Discrete Pneumatic Valves

Introduction . Pneumatic actuators are widely used in industry due to their reliability, simplicity of design, and ability to operate under complex conditions. However, when solving positioning problems, the use of traditional proportional valves is often redundant, which causes an unjustified increase in cost and complexity of the design. The application of simpler discrete distributors faces the problem related to the need to reach a compromise between their switching frequency and positioning accuracy. Existing studies mainly focus on optimizing individual performance indicators of pneumatic actuators and do not offer effective methods for finding a compromise between conflicting criteria. Using classical methods for constructing a Pareto set for multicriteria optimization requires significant computational resources, which complicates their practical application. The research objective is to develop a methodology for multicriteria optimization of the parameters of a positioning electropneumatic actuator with discrete distributors based on the construction of a Pareto set using surrogate models, which provides finding the optimal balance between switching frequency and positioning accuracy. Materials and Methods . The research was conducted on a model of a positioning pneumatic actuator with discrete distributors, implemented in MATLAB Simulink. The Latin hypercube method was used to analyze the parameters, which provided uniform filling of the parameter space. To reduce computational costs, surrogate models, built using neural networks, were used. Sliding control was selected as a control algorithm, which effectively compensated for external disturbances and uncertainties of the system. Results . The optimization of control parameters has shown the possibility of reaching high positioning accuracy with a minimum frequency of distributor switching. The use of the Latin hypercube method provided a uniform distribution of the calculation points, which made it possible to construct an accurate surrogate model. It has been experimentally proven that the proposed approach reduces computational costs by 48%, while maintaining high accuracy of modeling and analysis. Discussion and Conclusion . The research results confirm that sliding control is an effective solution for discrete pneumatic drives in the context of multicriteria optimization. The developed approach makes it possible to significantly reduce the frequency of switching distributors without substantial losses in the quality of transients, which helps to extend the service life of equipment and increase the reliability of automated systems. The use of surrogate models and neural network technology opens up new prospects for faster design of complex systems.