PSI - Issue 13

Mohammad Reza Khosravani et al. / Procedia Structural Integrity 13 (2018) 168–173

169

2 Author name / Structural Integrity Procedia 00 (2018) 000–000 data which show fracture of adhesive composite joints. Late r, in [2] the extended finite element method is utilized for analysis of fatigue life of the integral skin-stringer panel which are used in airframe construction. At the same time, Stojkovic et al. in [4] developed a mathematical model to predict strength degradation of composites which are subjected to constant amplitude fatigue. The researcher, refereed to the several experiments, and conducted stress analyses on di ff erent types of composite joints. However, literature investigations indicated that usually two kinds of failure prediction methods have been used for bonded composite joints in previous researches. The firs t ones are stress or strain based methods which use failure criterion equations. The second method is based on fracture mechanics approach which assumed initial cracks in the joint. The first method is not appropriated, due to the str ess or strain singularities in the composite joints. In the second method, crack growth assessment is necessary by comparing strain energy release rate with fracture toughness obtained from experiments. Non applicability of this method for bonded joints without an initial crack, and time consuming process of fracture toughness test, can be counted as demerits of this method. Beside of these criteria in mechanical engineering, di ff erent branches of artificial intelligence (AI) are employed to predict the response of composite to various loading conditions and estimate mechanical behavior and failure load [5, 6, 7, 8, 9, 10]. For instance, In [5] adaptive Neuro-Fuzzy modeling is used to predict life of the composite laminates. In this regards, experimental data used as input, and system predicts fatigue life. Later, in [8] fuzzy logic (FL) approach is employed to predict failure load of composite plates. To this aim, experimental results compared with results of fuzzy model. This comparison proved the capability of FL in prediction of maximum failure load. In [10] we reviewed applications of AI methods in fracture mech anics. In work [11] artificial neural network (ANN) are utilized to predict failure load of single lap adhesive joints. In this respect, the researchers used experimental data form the literature which includes various single lap adhesive joints under tensile loading. The achieved results from ANN proved its e ffi ciency in failure load prediction. Concerning the challenges in non-destructive evaluation (NDE) which are described in [12] some of the implemented AI systems can be considered as this type of evaluation. Although di ff erent AI approaches have been employed to predict failure load of composite joints, case based rea soning (CBR) is not utilized to estimate the occurrence of fracture, and predict failure load in adhesively bonded sandwich composite joints. CBR solves new problems by adapting previously successful solutions to similar prob lems. Compared with other AI techniques, CBR has some advantages, for instance it does not require an explicit domain model, and implementation is reduced to identify sig nificant features which describe a case. Moreover, a large volume of information can be stored, and a CBR system can learn by acquiring new knowledge as cases. Here, in the present study we utilize the CBR approach to implement an intelligent system which predict fracture incidence, and failure load in various sandwich-structured composite joints. The key idea of our proposed system is to overcome the di ffi culties of experimental tests. It is worth mentioning that the implemented system provides valuable informa tion about performance of the joints under di ff erent loadings. This information at the first stage of the joi nt design is beneficial to avoid time consuming experiments. The imple mented system is validated by using di ff erent experi mental data sets from literature which are detailed later. The achieved results are in a very good agreement with the experimental results and it’s showed capability of the proposed system. This paper is organized as follows: after this introduction, Section 2 present an overview of adhesively bonded sandwich-structured composite joints. Section 3 is dedicated to explain implementation of intelligent system. Section 4 shows obtained results and an evaluation of the system. Conclusions and future works are presented at the end. Combination of high strength and low weight materials, provided sandwich structures which are increasingly used in many industries. In the simplest form, two thin sti ff facing, which called skins, are separated by thicker core material. Technical assessment shows that sandwich struct ures have high flexural sti ff ness to weight ratio, lower lateral deformations, and high buckling resistance, so they play an important role in the improvement of composite structures. Generally, sandwich core materials are classified into two g roups: (a) homogeneous support of the skins, and (b) non homogeneous support of the skins. The first group can be made b y open or close cells, and the second one can be structured by punctual support, unidirection support, regional support and bi-directional support (honeycomb cores). Various types of materials have been used as core, face sheet and adhesive. In this study, we considered sandwich joints which are made by the following materials: 2. Overview of adhesively bonded sandwich joints