Article Preview
TopI. Introduction
When the ground is shaken strongly during an earthquake, some types of soils liquefy, often subjected to ground failures. Liquefaction is recognized as one of the main causes of earthquake-related ground failure. Different soil behaves differently in response to the earthquake. In case of saturated sandy soils, liquefaction occurs instantly. This phenomenon is very common in sandy soil. Liquefaction of soil is a major concern for any geostructure. Therefore, the determination of the potential for seismic liquefaction of soil is an important job in the geotechnical project. Several methods have been proposed to decide liquefaction capability of soil deposits. The stress-based procedure (Seed & Idriss, 1971; Seed et al., 1983, Liao et al., 1988) is the most general method practiced by engineers for liquefaction assessment. Idriss & Boulanger (2004, 2006) recommended revised semi- empirical strategies for getting to the liquefaction capability of soil. Juang et al.(2008) and Jha & Suzuki, (2009) carried out reliability study for liquefaction potential of soils utilizing SPT-test data. Samui & Sitharam (2011), A.Shahri (2016), Kumar & Rawat (2017) assessed and predicted the liquefaction potential using different artificial neural network models. Umar et al.(2018) have been employed the deterministic and probabilistic study of liquefaction for many areas in Bihar. The objective of this paper is to propose some model for soil-liquefaction potential classification. In latest years, many alternative techniques for computer-assisted pattern recognition have developed as a result of advances in computational software. The primary concept behind pattern identification systems such as neural networks, fuzzy logic or genetic programming is to learn from experience in an adaptive manner and to extract different distinctions. Artificial neural networks (ANN) are the most commonly used methods of pattern recognition used to determine the incidence of liquefaction based on SPT field data. ((e.g., (Hanna, Ural, & Saygili, 2007); (Juang and Chen 1999)). However, the major drawback of the ANNs approach is the network structure's large complexity. We are developing intelligent machines with successful applications. However, in order to improve machine learning, it is promising to imitate definite artificial emotions. This article introduces a neural emotional network based on the algorithm of emotional back propagation. Here the efficiency of an emotional neural network investigated. For the potential of soil liquefaction, the emotional neural network will be implemented. Experimental findings indicate that artificial feelings can be effectively modeled and applied to enhance the learning of neural networks and generalization. In scientific terms, a person's feeling at a specified time is called ‘emotion’. Researchers have studied the function of emotions in artificial intelligence from a multitude of perspectives. Human emotions have more than just a logical, reasonable element; they are closely associated with behavior and feelings. For the same reasons people will need emotions for future machines (Yoon, Sangjoo Park, & Anyoung Kim, 2004). Machines will need a type of emotion — machine emotion — when they have to operate continually without people's assistance. In human decision-making process, emotions play an important role. Abu and Zitar (2007) suggested and enacted an emotional agent design that resembles certain human behaviors. They noted that artificial emotions could be used to influence decision-making in different ways. While machine learning and making decisions on machines, we have always ignored the emotional variables ; however, it is quite conceivable to artificially model certain feelings in the learning algorithm of machine learning (EBP) (Khashman, 2008). An emotional neural network (ENN) is based on the emotional back propagation learning algorithm, which is an improved version of the conventional learning algorithm for back propagation. There are two simulated feelings in the emotional neural network that help the procedures of network learning and classification.