Techniques for the Tuning of Helicopter Multivariable Flight Control Systems and Handling Qualities

Helicopter flight control systems are often developed using low order linear descriptions of the plant. Unfortunately, unmodelled high order dynamics, such as those of the actuators and the main rotor, can have an adverse effect on stability and cross couplings when the design is tested on the aircraft. Hence, the flight controller may require tuning during commissioning trials in order to yield a system with acceptable handling qualities. As the sophistication of flight control systems is enhanced, the currently used trial and error optimization techniques will lose effectiveness. Anticipating the difficulties which will arise in the implementation of active control technology to helicopters, a study has been made of systematic procedures for adjusting the control system gains. The tuning processes which have been developed rely upon the signal convolution method to generate sensitivity functions of the state variables with respect to control system gains. State variable sensitivities allow one to predict what effects changing a controller gain will have on the system response. The beauty of the signal convolution method is that the sensitivity information is generated without knowledge of the helicopter plant. Therefore, by using data collected during flight trials, it is possible to calculate the sensitivity functions with respect to the dynamics of the actual system plant, including the unmodelled modes. The sensitivity information is used by an adjustment algorithm which employs Newton-Raphson techniques to predict how the system response will change with a trial set of perturbations to the controller gains. For each set of perturbations, an estimate is made of the modifed response which, in turn, is assigned a figure of merit. The set of perturbation values which yields the best figure of merit is then used to update the initial values of the control system gains. Since the characteristics of the optimized system response are determined by the type of figure of merit used in the adjustment algorithm, two distinct performance indices have been evaluated during the study. In model reference tuning, the Least Integral Error Square Performance Index is calculated to provide the figure of merit for each projected system response. The controller gains are altered to minimize the difference between the response of the actual system and a desirable response which is generated by a computer simulation model. However, in using a reference model, care must be taken to ensure that the desirable response is consistent with a Level 1 handling qualities rating so that pilots find the tuned system acceptable to fly. In contrast, the Handling Qualities Performance Index allows system responses to be compared explicitly in terms of whether or not they satisfy the handling quality requirements. As these requirements form the starting point for many control system designs, the use of the Handling Qualities Performance Index should guarantee an improvement in system response. This new performance index uniquely links the values of control system gains to the helicopter's handling quality ratings. Computer simulation has been used to validate both the application of the signal convolution method to multivariable control systems and the ability of the two performance indices to tune a helicopter's flight controller. The flight control systems considered during these simulations were developed using modal control theory and have been used with both linear and nonlinear representations of the helicopter plant. The results of a real-time simulation have reinforced the notion that the flight controller's structure and parameter values must be determined with respect to desirable flight handling qualities rather than purely on the basis of mathematical control system design techniques