A Fuzzy Neural Network for Fault Pattern Recognition

1　ＩｎｔｒｏｄｕｃｔｉｏｎＩｎｍａｔｈｅｍａｔｉｃｓ,ｆａｕｌｔｒｅｃｏｇｎｉｔｉｏｎｃａｎｂｅｓｕｍｍｅｄｕｐａｓａｍａｐｐｉｎｇｐｒｏｂｌｅｍｂｅｔｗｅｅｎｆａｕｌｔａｇｇｒｅ ｇａｔｅａｎｄｃｈａｒａｃｔｅｒａｇｇｒｅｇａｔｅ .Ｔｈｅｍａｐｐｉｎｇｂｅｔｗｅｅｎａｇｇｒｅｇａｔｅｓｉｓｃａｌｌｅｄａｍａｐｐｉｎｇｆｕｎｃｔｉｏｎ ;ｋｉｎｄｓｏｆｍａｐｐｉｎｇｆｕｎｃｔｉｏｎｓｃａｎｂｅｆｏｒｍｅｄｆｏｒｆａｕｌｔｐａｔｔｅｒｎｒｅｃｏｇｎｉｔｉｏｎ .Ｔｈｅｔｒａｄｉｔｉｏｎａｌｐａｔｔｅｒ…     

A quantitative identification approach for delamination in laminated composite beams using digital damage fingerprints (DDFs)

Digital damage fingerprints (DDFs) are a set of optimised and digitised characteristics of structural signatures, which are able to exactly and uniquely define a certain kind of structural healthy status. The DDF-based damage recognition technique includes the extraction of DDFs, assembly of damage parameters database (DPD) and subsequently inverse recognition in virtue of artificial intelligence. In this study, DDFs extracted from Lamb wave signals were employed to quantitatively assess delamination in carbon fibre-reinforced laminated beams. Characteristics of Lamb wave signals in the laminated beams were first evaluated, and DPD hosting DDFs for selected damage scenarios was constructed through numerical simulations, which was used to predict delamination in the composite beams with the aid of an artificial neural algorithm. The diagnostic results have demonstrated the excellent performance of DDF technique for quantitative damage identification.     

Application of a feasible formability diagram for the effective design in stamping processes of automotive panels

The objective of this study is to propose a method of process design that uses a feasible formability diagram, which denotes the safe region without fracture and wrinkle, for the effective and rapid design of stamping processes. To determine the feasible formability diagram, FE-analyses have been performed for combinations of process variables that correspond to the orthogonal array of design of experiments. Subsequently, the characteristic values for fracture and wrinkle have been estimated from the results of FE-analyses on the basis of the forming limit diagram. The characteristic values for all combinations within a whole range of process variables have been predicted through the training of an artificial neural network. The feasible formability diagram has been finally determined for all combinations of process variables. The stamping processes of automotive panels to support suspension module, such as the turret suspension and the wheel house, have been taken as examples to verify the effectiveness of process design through feasible formability diagram. A comparison of the FE-simulation results with the experimental results reveals that the design of stamping processes through feasible formability diagram is efficient and suitable for actual processes.