Polymers have been frequently employed in electrical applications because of their strong thermal and electrical insulating qualities, low density, and chemical resistance. In this study, a comparison between the behaviour and electrical properties of polymer blends and the results of artificial neural network (ANN) modelling has been conducted. Five samples of silicon rubber (SiR) and ethylene propylene diene monomer (EPDM) were prepared in different proportions. A dielectric test was used to test the dielectric performance of insulation samples under various polluting conditions such as dry, wet, low salinity, and high salinity wet according to ASTM standards. Percentage of blend and dielectric strength were used by ANN modelling for varying ambient conditions. The observations on ANN results and the experimental results have shown sufficient accuracy mutually. The artificial intelligence modelling studies for this article prove the applicability of the behavioural and electrical properties of EPDM/SiR blends. These findings indicate that artificial neural networks can be a useful tool for conducting experiments on the behaviour and electrical properties of polymer materials.

Because of the Variety in characteristics, strength, economy, and ease of manufacturing, the rubber blends are very suitable for use in the field of insulating materials. The blending technology has effects on the chemical, physical, mechanical, and electrical properties of polymers; this effect is predominately convenient for electrical insulation purposes. Electrical systems are often subject to faults resulting from short circuits or any other cause, which naturally leads to an increase in temperature for Insulating materials; And from here required considering good electrical properties and additional to have desired mechanical properties for insulation and bear it for different temperatures. The effect of thermal stress on the blending ratio of ethylene propylene diene monomer (EPDM) and silicone rubber (SiR) at various temperatures is studied using the Feed-Forward Neural Network (FFNN) after laboratory testing in this paper. The five different samples of EPDM-SiR blends (100/0; 75/25; -50/50; 25/75; 0/100) were prepared. The Breakdown Voltage (BDV) was measured under various temperatures (25, 60,100 and 130°C) according to ASTM standards. The experimental data was used to train the FFNN model. The blends ratio and temperatures represent the input of the FFNN system while the breakdown voltage kV is the output. The outputs obtained from FFNN were compared and checked against the data obtained in the laboratory. This study indicates that FFNN can be trusted to simulate the effect of thermal stress of various blending ratio on breakdown voltage with a satisfactory rate. It also demonstrates that the FFNN approach is an active tool that can be adopted as a reference to reduce the time and cost required in preparing and testing samples in the experimenter.