Similarity Measuring Approuch for Engineering Materials Selection

Advanced engineering materials design involves the exploration of massive multidimensional feature spaces, the correlation of materials properties and the processing parameters derived from disparate sources. The search for alternative materials or processing property strategies, whether through analytical, experimental or simulation approaches, has been a slow and arduous task, punctuated by infrequent and often expected discoveries. A few systematic efforts have been made to analyze the trends in data as a basis for classifications and predictions. This is particularly due to the lack of large amounts of organized data and more importantly the challenging of shifting through them in a timely and efficient manner. The application of recent advances in Data Mining on materials informatics is the state of art of computational and experimental approaches for materials discovery. In this paper similarity based engineering materials selection model is proposed and implemented to select engineering materials based on the composite materials constraints. The result reviewed from this model is sustainable for effective decision making in advanced engineering materials design applications.


Introduction
Engineering materials are the artificial materials, such as Polymer, Ceramic, Metal and their composite with fiber reinforced materials, which are being used in our daily life. Any two materials could be combined to make a composite and they might be mixed in much geometry. Selection of design and fabrication processes associated to engineering materials design is the tedious task that is being faced by the most of the manufacturing industries. The selection of appropriate materials, which meet the design requirements and improve the performance, reliability, durability of composite material, is the critical task in Computer Aided Design (CAD) and Computer Aided Manufacturing (CAM) systems [5].
As wide variety of more than 50000 materials available today and varying in their characteristics and costs, materials selection system is very much essential to ease the difficult complex process. This selection process involves decision-making strategies in determining the prerequisite materials that suit the design specifications and requirements of composite design.  [12]. The applications of expert system play major role in diverse application fields from materials design and their manufacturing. Design of computational expert systems on wider range of data sets have still research scope in advanced engineering materials design applications [6] [13][14]. Therefore, Composite Materials Selection System (CMSS) is proposed and implemented in this paper.
The paper has been organized as follows. The second section presents the composite materials selection system. The third section describes similarity measure functions. The forth section describes the selection strategy on different materials type. The last section concludes the work and briefs the future work scope.

Com
Expert systems are programs in whic knowledge about a problem is embedded in a set of modules called as rules, frames, objects, or scripts that are stored in a repository called a knowledgebase. The Composite Materials Selection System (CMSS) is developed in order to simplify the complex selection process for opting appropriate materials that meet the design requirements. The structure of the proposed system is shown in the figure 2.
The CMSS consists of several integrated modules are responding for potential input parameters to produce outputs that are treated as inputs of another module. The integrated modules of CMSS are input module, Indexed based classifier [9] [10], fragment database generator, distance measure computation module and materials selection module. All these modules are simplified with non-redundant computational effort.
The input module (l e CMSS a list of materials characteristics that are specified by design engineers. It will be interacting with both the indexed based classifier and fragment database generator. Index based decision classifier scans through the inputs and segregates materials characteristics/ attributes into different classes that are represented by nodes. The segregation of attributes into different classes based on the classification rules defined in the knowledgebase of the system. The outcome of index based decision classifier is forwarded to the fragment database generator that selects the portion of the database containing matching attributes with the tuples belonging to materials class as predicted by the index classifier.

Composite Design
The composite design specifications ar [6] of a component to be designed and a design engineer derives these parameters. Design requirements are the properties of primary importance such as physical properties, mechanical properties, chemical properties, thermal properties and so on. These properties represent quantitative attribute and linguistic values of a component. There are 23 properties considered in this system. Some quantitative properties have range values (Density: 0.23cm 3 to 0.56cm 3 ) and others properties have ordinals/linguistic/categorical values (Poor-Excellent). Each ordinal/linguistic value is replaced with a unique numeric weight. nowledgebase [7] is defined as "A database of subject; used in Artificial ision classier shown in figure 3 is a that is used as decision- In the first step of classifier, when a property is randomly sampled from a design requirement list, the classifier invokes the rules defined in the knowledgebase and creates a node in the class corresponding to the index pattern.

Generating Fragment Database
The Material Database (DM) stores all classes of materials, C = {P, C, M}. Each class is having the materials attributes,

Knowledgebase
K knowledge about a Intelligence. The knowledgebase for an expert system (a computer system that solves problems) comes partly from human experience and partly from the computer's experience in solving problems. It must be expressed in a formal knowledge representation language for the computer to use it". The knowledgebase of CMSS consists of 23 decision rules and each decision rule generates a prime index pattern that represents a material class.

Index Based Classifier
Index based dec simple and robust classifier making principles in most of the fields such as Machine Learning, Pattern Recognition, Image Processing and Data Mining and Knowledge Discovery. It discriminates design requirement properties based on the expert rules defined in the knowledgebase. Each class generated by the classifier is implemented with linked lists. Each node in a list has three fields including Property Name (PN), Property Index(PI) and a Pointer(Ptr) for respectively storing the next property name as defined in the input design requirement list, index value generated by the decision classifier, and the next node address.
A class of materials fra ase is proportional to O(N) time complexity in the best, average and worst cases of analysis. This fragmented data space reduces the computational efforts with less memory space during computing distance measure gmented from material datab values.
The resultant database obtained by (1) The distance is symmetric an object, x, to an object, y, in space is no more than making a detour over any other object other than object z (triangular inequality).
There are various popular distance measuring functions that are satisfying the above principles. Euclidian distance measure [7] metric is employed for distance computations. This distance measure metric is defined as follow:  (4) iii. Absolute Exponential measure: (5) . Geometric Average Minimum: (6) v. Correlation  = dimensional data objects. a object space for an input data object. The best match data object is determined by the Euclidian ce computation. This has been using as standard istance measure function in data mining and edge discovery [7]. The best match object for an object is selected through the determination of the are two n

Similarity Material Selection
It is t e process of selecting the best match dat h in data for an input distan d knowl input least similarity measure value.  figure 4. Initial selected class are computed. A material corresponding to the least distance is selected as the potential material that meets the design requirements.
Design specification specified by design engineers are the input parameters that through classification on input design requirements into Polymer, Ceramic and Metal Classes, the fragment materials data sets generated and the material selected by the Euclidian measuring technique from the different classes are shown in the figure 5.     7) and (8) belonging to L 1 family and are competent enough to select the materials that are very closure to the input specification. However, the function (6) and (7) are feasible for materials selection, but function (7) is more appropriate for analyzing redundancy and consistency among the materials data sets. Function (6) is not the feasible one as it maps to the different material in the class.
The L 1 family functions and the functions (6) and (7) are compared and shown in the Table 2 and their performance evaluation on numeric approximation is depicted in the figure 7. The degree of similarity of Euclidian distance function is less that emphasizes much of functions (6) and (7) depicted in the table 2 and in the fragmented data sets are shown in the table 2. From this table 2, it shows that distance/similarity closeness among the L 1 family functions. The degrees of similarity ( Sl. nos. 4 and 5 ) are still less than the Euclidian distance measure, however, one of these functions (6) maps the input design requirements to the materials that