An Application of Deep Learning Using Leaky Rectified Linear Unit and Hyperbolic Tangent in Non-destructive Testing

An artificial neural network (ANN) is a data processing technique where a set of hidden layers consists of several neurons complexly interconnected. ANN can find patterns between source data and response data and correlate them using an algorithm to predict the desired response. This study is concentrated on the development of mathematical models and error matrices using the activation functions such as Leaky Rectified Linear Unit ReLU and Hyperbolic Tangent (Tanh) used in the hidden layer and output layer respectively from the weight and biases extracted from MATLAB NN tool. The activation functions were employed on weight and biases extracted from ANN simulation and the error matrices were then obtained by interchanging their positions in the hidden and output layer. This was performed by backpropagation with one hidden layer neural network adopted as ANN structure to predict the compressive strength of reinforced cement concrete (RCC) using the non-destructive technique (ultrasonic pulse velocity).