International Journal of Engineering Research and Advanced Technology (IJERAT) E-ISSN : 2454-6135 DOI: 10.31695/IJERAT.2019.3429 Volume.5, Issue 4 April -2019 www.ijerat.com Page 19 Licensed Under Creative Commons Attribution CC BY Modeling of Peak value in Cylinder Firing Pressure in Diesel Engines using Artificial Neural Network Methodology Ozan Doğukan BOZDAĞ 1 and Orkun ÖZENER 1 1 Department of Mechanical Engineering, Internal Combustion Engines Laboratory, Yıldız Technical University, İstanbul,Turkey ______________________________________________________________________________________ ABSTRACT From the past to the present, restrictions have been imposed on emissions due to the widespread use of internal combustion engines and the environmental damage caused by emissions. Engine manufacturers perform calibration studies by affecting the fuel-air parameters through the engine control unit to ensure emission limits. The maximum pressure due to the inside of the cylinder is called the maximum pressure inside the cylinder. In order for the engine to work properly, there is a cylindrical pressure resistance due to the material strength. When performing calibration work, the maximum pressure inside the cylinder should be monitored and maintained within the strength limits. In-cylinder pressure sensors are used in the in-cylinder pressure monitoring. Because the pressure sensors are exposed to high pressure (more than 200 bar or more in heavily heated), the values they read can be reduced or distorted. In this study, 6 different the artificial neural network (ANN) model design was created and trained with 2400 test points obtained from engine dynamometers system. Then these networks are tested again with 60 unused test points which are not used during training phases. Then the results are analyzed in terms of peak firing pressure difference. The research results showed that 2 neurons ANN system is best ANN system capable of predicting peak firing pressure within 1.7 bar average difference to measured data. Key Words: Diesel Engines, Artificial Neural Networks, Peak Firing Pressure In Cylinder, Experimental Study, Engine Calibration ______________________________________________________________________________________________ 1. INTRODUCTION In spite of the increasing torque and performance requirements of diesel engines, the reduction in emission limits and the improvement in fuel consumption put a great deal of burden on engine manufacturers. The torque, performance and fuel consumption, which are directly proportional to the efficiency of combustion, increase the thermal and mechanical loads in the cylinder. During the operation of the engine a pressure rise occurs due to the internal combustion process. During this pressure rise, the maximum pressure is called the maximum internal pressure within the cylinder. One of the conditions required for the engine to continue to operate in a healthy way is that no pressure value during combustion exceeds the maximum pressure resistance limit of the cylinder. The maximum pressure in the cylinder reaches up to 200 bar (even higher on heavy commercial vehicles). The higher the maximum compressive strength of the engine in the cylinder, the more efficient combustion will be achieved and the amount of torque taken from the engine will increase [1]. During the engine development tests, the maximum pressure inside the cylinder must be monitored. When these values exceed the resistance, actions must be taken to protect the engine. Today, maximum pressure values are followed by engine cylinder pressure sensors. These sensors are usually installed by means of an adapter instead of the glow plug. When both sensors and glow plugs are needed, the engine head must be machined. This process will affect the internal turbulence of the cylinder. On the other hand, there are disadvantages such as drifting as a result of the use of sensors, high costs due to high pressure, installation of one cylinder per cylinder, use of amplifier, module and computer [2]. In this context the peak firing pressure (PFP) data is monitored via developed computer algorithms to prevent from damage that arouses from sensor faults and etc. In the automotive sector, there lots of modelling techniques that are used. Some of them are using massive and complex algorithms. But these modelling techniques are generally needs more computing time for calculation. Considering to the diesel engines maximum rotation speed (i.e. 6000 rpm) there is 3000 PFP peaks are created and all of these peaks should be monitored in a predefined accuracy. The ANN approach is a new, fast calculation methodology that does not require complex mathematical equations to explain a non-linear and multidimensional system. Over the last decade, more attention has been paid to ANN techniques, particularly in the