Architecture of a cyber-physical system for aircraft fuel control systems tests

The aircraft fuel control system makes sure an engine has the precise amount of fuel required to generate power and thrust for every stage of flight. It therefore plays a role in determining the economics of flight. In order to maintain the economic equilibrium of performance and flight safety, it undergoes a Maintenance Repair or Overhaul (MRO) service after several hours of operation. A critical aspect of the MRO service is the test performed to validate the airworthiness of a fuel control system before it returns to service. The test of aircraft fuel control systems is human-centric by design. The operator uses a network of test systems to generate in-flight conditions similar to what the Unit-Under-Test (UUT) experiences on-board an aircraft engine, then performs tests to validate and verify its airworthiness. Thereafter, test results are recorded for regulatory compliance reasons after each test is performed successfully. An analysis of the test specification for the UUT involved in this study revealed 90% of an operator’s touch-time is automatable. The functions of control, data processing, data entry and supervision must be achieved automatically if they are to be performed autonomously by a cyber-physical test system. But the automation of these activities at the micro level does not guarantee their autonomous execution at a macro level by such a cyber-physical system composed of the network of test systems. Therefore, knowledge of a multidisciplinary array of fundamental concepts and how they can be fused to execute the test of aircraft fuel control systems autonomously, have been developed as presented in this thesis. For the function of process control, the response of the processes used to set test conditions is ~ 50Hz, five times the rate of process responses reported in typical process industries where automation of process control have been achieved. As a result of this fundamental knowledge, the design of the architectures for the functions of control and data processing is an asynchronous one. Noting that none of the data is fed back through a network like the case of Networked Control Systems or Supervisory Control and Data Acquisition Systems. The realization of the control functions for each process used to set test conditions is based on control laws synthesized through modelling of their respective actuation mechanisms. Of the three models developed, a 2nd order model has been identified as been representative of the dynamic and steady-state characteristics of each actuation mechanism. A typical actuation mechanism contains a high number of masses and springs whose physical modelling resulted in a model of 12th order. This model is highly unrepresentative of the transient and steady-state response observed in the process due to difficulty in estimating the internal parameters of respective actuators. A linear model synthesized from the calibration data of each actuation mechanism has also been investigated and found to be too ideal. Its response is unrepresentative of the dynamic characteristics of the actuation mechanisms. The processes used to generate test conditions have been set simultaneously using a network of PID controllers. The controllers’ gains are an order of ten less than what they were for the sequential set up of test conditions due the fact that there are interactions between the processes inside a UUT. Fundamentally however, this Thesis demonstrates an asynchronous architecture for the control function, which enables a pseudo steady-state execution of tests. In-practice, this has the potential to reduce the time it takes to perform a test by one-third. The function of supervision has been developed in the form of a Fault Detection Isolation and Recovery capability within the cyber-physical test system. The architecture for this function is designed based on minimalizing the constraint of the period of recoverability (PoR), where the deviations from normal operation need to be detected, the outcomes they could result to—diagnosed and recovery strategies executed to prevent test systems or the UUT deviate from normal operation, using measurements acquired in a time