Failure Modeling of C-130 Turbines using Artificial Neural Networks Nizar A. Qattan, King Abdulaziz University Ali M. Al-Bahi, PhD, King Abdulaziz University Belkacem Kada, PhD, King Abdulaziz University Key Words: Reliability, Neural Network, Back Propagation Algorithms, Turbine Blades SUMMARY & CONCLUSIONS The C-130 aircraft is one of the most widely used medium transports in the world. It operates virtually everywhere, from the arctic circle to the Sahara. Operation in desert conditions, however, presents a challenge for maintenance engineers regarding preventive maintenance scheduling. Erosion caused by sand particles drastically decreases turbine blades life. Recent studies showed that Artificial Neural Network ANN algorithms have much better capability at modeling reliability and predicting failure than conventional algorithms. In this study, more than thirty years of local operational field data were used for failure rate prediction and validation using several algorithms. These include Weibull regression modeling to establish a reference, feed-forward back-propagation ANN, and radial basis neural network algorithm. Comparison between the three methods is carried out. Results show that the failure rate predicted by both the feed-forward back-propagation artificial neural network model and radial basis neural network model are closer to actual failure data than he failure rate predicted by the Weibull model. The results also give an insight into the reliability of the engine turbine under actual operating conditions, which can be used by aircraft operators for assessing system and component failures and customizing the maintenance programs recommended by the manufacturer. 1 INTRODUCTION Modern aircraft engines are very complex machines. They provide the necessary thrust for the aircraft to fly. Therefore, the safety of an aircraft greatly depends on the reliability of its engines. The extreme high temperature, pressure, and velocity of the intake air mass may contain sand and dust which will cause a catastrophic damage to aircraft turbine and engine. So preventive maintenance and continuous monitoring of engines are essential measures to increase both reliability and aircraft safety. The Turbine system is a 4-stage turbine designed to extract the air energy directed from the combustion chamber at extreme high pressure and temperature - maximum Turbine inlet temperature (TIT) of 1077ºc at Take-off power limited to 5 minutes, 1010ºc maximum continuous operation and, 932ºc recommended cruise power - develop 11000 Hp of mechanical energy to drive the compressor, propeller, and engine accessories. As we mentioned in the introduction part, the turbine section is the most effected by thermal distress, sulfidation and sand ingestion. The Turbine system consists of many components, the man turbine components are: Turbine inlet casing, vane and seal support, Turbine vane casing, four stages of turbine stator, four stages of turbine rotor, Thermocouples and rear bearing support, as presented in Figure 1. To simplify our modeling, we will deal with the engine turbine as a single unit. Figure 1 Turbine Unit Assemblies Artificial Neural Networks were introduced several decades ago as a means of modeling. There have been numerous publications explaining these methods, see for example [9, 10 and 11]. The convergence criteria were reduction of mean square error to a minimum value. This delta rule for a single layer can be called a precursor of the back- propagation net used for multi-layer nets. The multi-layer extension of Adaline formed the Madaline. In 1982, John Hopfield’s introduced new concept networks, Hopfield showed how to use “Using spin glass “type of model to store the information in dynamically stable networks, [1]. His work paved the way for physicists to enter neural modeling, thereby transforming the field of neural networks. Three years later, Parker back propagation net paved its way into neural networks, [2]. This method propagates the error information at the output units back to the hidden units using generalized delta rule. This 978-1-7281-8017-5/21/$31.00 ©2021 IEEE 2021 Annual Reliability and Maintainability Symposium (RAMS) | 978-1-7281-8017-5/21/$31.00 ©2021 IEEE | DOI: 10.1109/RAMS48097.2021.9605792 Authorized licensed use limited to: Nizar Qattan. Downloaded on June 30,2022 at 13:21:41 UTC from IEEE Xplore. Restrictions apply.