MODELING OF RECONFIGURABLE MANUFACTURING SYSTEM WITH AVAILABILITY CONSIDERATION

Abstract

The last few decades have witnessed fierce competition and dynamic business environment in manufacturing sector due to globalization of economics, fast pace of growth in technology and customer driven market scenario. For manufacturing companies to be competitive in this global market, they must have three foundations namely low cost, high quality and rapid responsiveness towards change. Low cost has been an economic goal since the introduction of mass production. High quality is also an economic goal when focus on six sigma by improving the system operations. Cost effective rapid responsiveness can be achieved, through reconfiguration, with focus on new ways of designing and operating the manufacturing system. This gives motivation to researchers to think differently and subsequently a new manufacturing paradigm has emerged which is known as Reconfigurable Manufacturing System (RMS). Reconfigurable manufacturing system is capable in responding quickly whenever changes are required in functionality (adding/removing machines/modules/tooling in system) and in capacity (increasing or decreasing production volume) as per demand. The maneuvering of functionality and capacity of the system according to needs is providing the benefits of mass production and at the same time is giving flexibility to handle the variety of products. RMS possesses six core characteristics. Out of these six, four are considered to be as primary and remaining two are considered as secondary. The primary characteristics are customization, modularity, convertibility and scalability. The first two primary characteristics are related with the economics while remaining two primary characteristics are related with the rapid responsiveness of the system in terms of capacity and functionality. The secondary characteristics i.e. integrability and diagnosibility are related with the control, detect and diagnose the root cause of defects during operation and subsequently correct them quickly. Therefore, these secondary characteristics act as supporting characterstics for reconfiguration. For demonstration and justification of all the promises made by this new manufacturing paradigm, there is need to develop the new and innovative approaches for the modeling of reconfigurable manufacturing system. The highlight of the work carried out in the doctoral thesis has been summarized in the following: v The success of a RMS is dependent upon the formation of optimum set of part families addressing the various reconfigurability issues. For this, a three phase methodology is developed to form part families on the basis of similarity among parts for a demand period in consideration. In the first phase of methodology, a mathematical model is developed to find the operational sequence similarity among the various parts in considerations. The developed mathematical model results a correlation matrix. The resulted correlation matrix is a similarity matrix, shows the similarity among parts on the basis of operations sequence. The similarities among the parts are used form the group of parts which is achieved in the second phase of the methodology. In this phase of methodology, Principal Component Analysis (PCA) is used to form the group/cluster of parts on the basis of similarity among the parts. The first principal component extracts, in principal component analysis, the maximum variance among the operations sequence similarity. This means that the first principal component is correlated with at least some of the part’s operations sequence similarity. The second principal component extracts variances which is not associated in first principal component. The output of this analysis provides a scatter plot in which each dot is representing a part. Closely packed parts show high correlation. Further, to sustain our claim of correlation, an analysis is done on the basis of angular distances measure among the parts with the help of scatter plot. The above analysis provides only clustering (grouping) of parts and not any level for similarity which is achieved in the third stage of developed methodology. To do this, an Agglomerative Hierarchical Clustering Algorithm is modified and used. The modified algorithm computes the Euclidean distance among part clusters/ groups formed in second stage of methodology and store in it. Parts cluster having minimum distance with other part clusters is clubbed first and the clubbed cluster is then removed from the next iteration and so on till last cluster clubbed. The output of the algorithm gives a dendrogram, shows the similarity level. An example case has been taken to demonstrate the developed methodology. The developed methodology provides better results when we compare the results of same example problems solved by another well known clustering algorithm (k-means algorithm) and this validates our claim. Previous section describes the formation of part families for RMS. It is developed on basis of similarity criteria and obtained a hierarchical clustering of parts. These hierarchical clustering methods always yielded a dendrogram. Dendrograph is a tree like structure which represents family of parts and similarity level for changes in part families accordingly. Initially system is configured to produce the first selected family and then it vi is reconfigured to effectively produce the following part family, and so on. Whenever there is change in the configuration of the system, the firm incurs a changeover cost which considers the present configuration and the targeted configuration. In a particular case if all parts group in a single family, then system requires all the necessary reconfigurable machine tools for production of parts. This particular condition leads toward the situation in which the changeover is minimum but the system is having maximum underutilized resources. This is due to the more number of idle machines and the functionality of the system is not fully utilized. On the contrary, if the selection is such that each part is itself a family, than system requires the frequent changeover which incurs more associated changeover costs. Although in this case, the number of idle machines is minimum and but their capacities and functionalities are fully utilized. To resolve the conflicting situation, a mathematical model is developed to select and find the optimal sequence for all part families in consideration of reconfiguration and underutilization cost. This is achieved by calculating the reconfiguration and underutilization cost at each level of the hierarchy. Reconfiguration cost constituents the cost of adding or removing a single module from a machine, cost of removing/ adding single machine from the system according to the requirements for manufacturing. While under-utilization cost constituents the cost of not utilizing the resource module in machine. It is calculated per unit time and incurs due abundant resources in the system. Using the definition of cost parameters mentioned above, the selection and sequencing of the part family’s problem give the minimum cost solution by solving the Travelling Salesman Problem (TSP) at each level of dendrogram. The developed mathematical model for sequencing of part families is solved by ACO technique and achieved the optimum solution. Typically, Reconfigurable manufacturing system consists of the several reconfigurable machines having several finite performance levels depending upon the different state of the machines. Accordingly, if a system having a finite number of the performance states can be fall under the category of the Multi State System (MSS). For evaluation the availability of multi-state system, Universal Generating Function (UGF) technique is fast, efficient and easily adoptable. It is based on the algebraic recursive technique which produces the excellent result for large size multi state system. The application of UGF in assessing the availability of reconfigurable manufacturing system is developed. A modification in the original method is developed and expanded it to be used in reconfigurable manufacturing system capable of producing multiple part types simultaneously. For illustrating the above approach, a reconfigurable manufacturing vii system that produces multiple part types simultaneously is considered to explain the application of the modified UGF technique in evaluating system availability and its computational merits. Multistage system provides several operational configurations. Symmetric configurations are suitable for reconfigurable manufacturing system whereas asymmetrical (not symmetrical) configurations are not feasible due to immense complexity in material handling. The configuration of a system helps / hampers the system’s productivity, convertibility, scalability and responsiveness. Each alternative configuration provides a typical performance in these areas. Mean Time to Repair (MTTR) and Mean Time to Failure (MTTF) is measured to obtain the availability of the constituent machine/s of the system which further used to determine the productivity of configurations by Universal Generating Function. The system capability to rapidly change in the production functionality by switch over from one product to another product is defined as convertibility metric measure. The ability to precisely change the manufacturing capacity of a system through system reconfiguration with minimal cost, in minimal time, over a large capacity range, at given capacity increment is defined as Scalability. A high scalable system requires less lead time when a change in capacity is necessitated. Cost is the final performance metric that is always considered. A methodology is developed to select the system configuration based on the performance on these core characteristics for reconfiguration. An example case has been solved to demonstrate the developed methodology. A detailed performance analysis is done on the example case to get the pertinent data. Because of simultaneous consideration of several criterions, Entropy-based Analytical Hierarchy Process is used for the further analysis. On the basis of the evaluation carried out, a selection decision is made. Finally, sensitivity analysis is done to confirm that the selection results are robust.