Applying Bigdata on Specific Phase of Production Lifecycle Management

...

In this paper we will discuss and provide a framework that will be helpful in implementation technique and handling the Big Data in each phase of product lifecycle management (PLM). In a paper “Big Data in product lifecycle management”[2], researchers briefly descriebed the phases of PLM as BOL, MOL and EOL. Comparing with their work we will give a framework that how we can implement Big Data techniques and handle that data in more effiecent way. For this purpose we will combine two different Big Data algorithm C4.5 and K-means. In C4.5 algorithm work in form of decision tree and K-means algorithm it creates groups of simillar data sets. In each specific phase of PLM first we will make a data sets of similar data sets using K-means algorithm and then decision tree will be implemented through C4.5 algorithm.

-

Methodology

During the research we used two existing “Big Data” algorithms C4.5 and K-meas algorithm. Before disusing the methodology used in phases of PLM lets discuss the working of these two algorithm.

-

C4.5 Algorithm

C4.5 Algorithm is used for generating the decision tree.This algorithm is developed by Ross Quinlan. This algorithm is extension of ID3 algorithm which is also developed by Quinlan. The decsion tree which is generated in C4.5 can be used for classification. C4.5 build decision tree from a set of training data in a same ways as ID3 by using the concept of entropy. The training dataset is set of previously classified samples. For example S=s1,s2,s2…. In C4.4 data is stored at each node same as ID3. At every node of the tree, C4.5 picks the attribute of the data that splits its set of samples into subsets which enhanced in one class or more. The splited data is normalized. After that the attribute which is highly normalized is picks up for the decision .C4.5 work as follows:-

First it checks the base case then for each attributes it find the higher normalized data for spliting then its create a decision tree on the basis of high normalization at the end call the sublist recursively and then add nodes as children nodes[5].

-

K-means Algorithm

K-means algorithm is used for making k groups of similar dataset. It is a popular technique of cluster analysis to examine a data set. It is developed by James Macqueen through the idea proposed by Hugo Steinhaus. K means algorithm makes the group of similar datasets which are the nearest classified neighbor. In this algorithm the main idea is to exemplify the k centroids for each of the cluster. This centroids then placed in a smart manners because each distinct location cause distinguish result. After that each belonging point of given data sets it associate to nearest centroid. When there is no point left the step is completed and groupingin done. After tha recalculate the K-new centroid. Repeat the groupage steps until it produces a new group[4].

As we discussed in the previous section that there are three phases of product life cycle (BOL,MOL,EOL). All the related data to PLM is stored in databases.In BOL period our main focus is on marketing analysis and product design. Where we are looking for promising customer and their need towards the specific product.So we can gather datafrom past marketing analysis, from questionare, from user how needs that producr and by surveys. In MOL period the main phases are warehouses managing, product transport,customer services,preventive and predictive maintenance. Now a days because of huge data in warehouse “Big Data” techniques are required. If warehouses are advancing then transferring the product is also a big deal because of smart cities. Because of all this customers are increasing day by day so providing them a proper services again “Big Data” techniques are required, and similarly so for their preventive and predictive maintenance. At the end in EOL period product recovering decision making we can gather a huge amount of data on the basis of customer feedback,predictive and real time analysis of product and the remaining time of the product. All that data from these each specific phases are stored in databases with concepts of “Big Data” techniques like Hbase,Hadoop and many others. For that we mine out all the use full data through data mining algorithms and then make the sets of similar data sets with the help data clustering algorithms. For this purpose we are C4.5 for mining the data and K-means for data clustering. When we got the data then we can make similar data sets and apply tree decision algorithm to make the best decision. According to our methodology in this paper we combined two algorithms and proposed a way to handle specific phases of PLM. With the help of this we can handle the data more precisely.

-

Advantages of C4.5 and K-means Algorithm

- C4.5 frames models that can be easily be explained

- C4.5 can be implemented easily

- C4.5 can use absolute and continuous values

- If variables are huge, then K-means make the computations faster then hierarchical clustering.

- K-Means crop strong clusters than hierarchical clustering espically in the case when cluster is global.

-

Analysis

In our research we gather our overall data from the past findings about related products and user experince and feedback. We try to anaylis the draw backs of the past products and try to figure out what user need in what design the product could be more feasible. According to our research we try to make the groups of similar data sets and then make decision tree. We use these methods because our basic purpose is to predict and make decision base analysis that what could be better for the product. Before the usage of “Big Data” techniques in PLM the product growth is not much as compared with the usage of “Big Data”. As the data is growing day by day in PLM phases “Big Data” techniques are essential. Growth of “Big Data” and Success of using “Big Data” in PLM are shown is figure below.

[pic 1]

[pic 2]

-

Conclusion

In this paper the first contemporary improvement is revealing the specific phases where