PSI - Issue 13

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

171

4 Author name / Structural Integrity Procedia 00 (2018) 000–000 in performance and stability of the composite sandwich joints. Some of these parameters and related reference are presented in Table 1. It should be noted that beside of the mentioned parameters in Table 1, we considered specimen sizes, adhesive type and thickness, test environment and type of loading (e.g. cross-head movement). Therefore, as shown in Fig. 2, the case base is created based on the cases generated with these parameters. According to the impact value of each parameters in failure mode and based on the doma in expert consideration, we have assigned significant weight for each parameters. Thus, we can represented our cases with these parameters and weights or attributes-value.

Fig. 2. CBR overview of the proposed system.

Definition 1. (i) An attribute-value description is a set vector of attributes. (ii) For a given set E of experimental parameters, an attribute P assigns to each test T ∈ E some value taken from the domain of P . (iii) A case base ( CB ) is a collection of cases. After determination of case representation, the main phase in our methodology is case retrieval. Suppose we have a new test as a query problem and a defined case base to select mo st relevant experiments from. Therefore, we need a similarity measure to assess similarity between the new incoming case with the cases stored in our case base by finding the similarity between attributes or relative relev ance of each attribute. Definition 2. CB has a set of experiment parameters and q be an object from query; then some p of CB are more similar to q to find a nearest neighbor to q . Our local-global similarity measure is as follows:

w i j sim ( p t , p j ) m

S IM ( nt , C i ) = Σ k

m

(1)

t = 1 Σ

j = 1

where w i j is the weight of attribute j in the i th case, m is the number of parameters (attributes), k is the number of query’s parameters and sim ( p j , p t ) is the similarity (local) between parameter j ( p j ) from case i and parameter t ( p t ) from the new test (query). The last step in our CBR system is case reuse to make use of the retrieved solution by recommending its failure load and type of failure appropriately.

Table 1. Some stored parameters in the case base Ref. Type of joint Core material

Core thickness ( mm )

Skin thickness ( mm )

Joint length ( mm )

Skin material

[20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

T-joint T-joint L-joint T-joint Π -joint T-joint T-joint T-joint

PVC foam Balsa wood

14 38 15 30 13 50 15 20 4.5 6

CFRP GFRP GFRP GFRP GFRP CFRP GFRP CFRP GFRP CFRP

1 2 5 1 1 2 2

32

250 250

Nomex honeycomb

0.24

PVC foam

40 50 10 25 25 20

Nomex honeycomb Nomex honeycomb

Insert joint

PVC foam PMI foam PVC foam

2.1

200

Insert joint

Aluminum alloy

0.5