Turboshaft engine steady-state open-loop control design method based on machine learning algorithm

Abstract

The invention discloses a turboshaft engine steady-state open-loop control design method based on a machine learning algorithm. The turboshaft engine steady-state open-loop control design method usestwo PID controllers to form an inner loop and an outer loop to control the rotating speed of the turboshaft engine, so as to obtain input and output data of each steady-state point of the turboshaft engine, can restrain interference falling into the inner loop in time through the gas turbine negative feedback inner loop, can reduce the influence on the outer loop, and is conductive to the outer loop to overcome load interference falling into the outer loop, so that the whole system has high-quality control performance. The turboshaft engine steady-state open-loop control design method dividesthe acquired steady-state point data into two groups, one group being a training sample set and the other group being a test sample set;

Description

technical field [0001] The invention relates to a method for designing a steady-state control law of a turboshaft engine based on a machine learning algorithm, and belongs to the technical field of aviation turboshaft engine control. Background technique [0002] The control system of an aero-engine is a very complex and nonlinear control system. Within the flight envelope, its model is time-varying and uncertain. When the operating conditions change or the flight conditions change, if the aero-engine is not controlled , it will be difficult for the aero-engine to work normally, which requires monitoring the parameters of the engine during the entire flight and adjusting the control amount to make the parameters of the engine change according to the normal law to meet the requirements of the flight conditions. Precise and fast control of aero engines. Therefore, aero-engine control has always been a field of continuous exploration and innovation by researchers. [0003] Av...

AI Technical Summary

Problems solved by technology

[0006] The traditional controller needs a long response time to iteratively calculate the fuel supply to the model, and it is difficult to ensure that the engine reaches a steady state quickly, which directly affects the real-time performance of the helicopter, while the support vector machine is useful for thousands of data. Very good fitting ability, the final training of the standard support vector regression machine comes down to a quadratic optimization problem, and its training cost is relatively high, which can also be handled for small training sample sets, but for thousands of sample sets, It is quite difficult to deal with it on a computer using the effective set method

Images