Supplier-based Manufacturing

Manufacturing is a cornerstone of the U.S. economy. It is an integral part of a web of inter-industry relationships that create a stronger economy. According to the Bureau of Economic Analysis, every dollar of final demand spent on a manufactured good generates $0.55 of GDP in the manufacturing sector and $0.45 in non-manufacturing sectors. The economic downturn started in 2000 hit the manufacturing sector particularly hard. Today, as the overall U.S. economy expands strongly, much of the manufacturing sector continues to lag behind by a wide margin. The challenges faced by the manufacturing industry are mainly structural, resulted from the effects of rapidly changing technology and adjustment to a global economy. To meet these challenges, original equipment manufacturers (OEMs) are increasingly becoming outsourcing oriented. They view themselves as system integrators managing suppliers and distributors to meet customer demands. As a result, manufacturing competition has shifted from a company orientation to a supply chain orientation. Therefore, competitive advantages must be built at the supply chain level.

Our research centers around a closed-loop process design philosophy to virtually integrate OEMs’ design and suppliers’ manufacturing activities, as shown in Figure 1. It works as follows. First, process design alternatives are generated based on product design. These process designs are simulated to predict the resultant part quality given a supplier’s process capability. Trade-off among different aspects of part quality is then evaluated quantitatively to identify the best process design. The simulation and evaluation results can be used to provide feedback to improve product design and to identify supplier process bottlenecks. These bottlenecks are then eliminated through process optimization, leading to improved supplier process capability.

Figure 1. Closed-loop process design.

Process Design Generation

We have developed a computer-Aided manufacturing planning (CAMP) methodology to generate process designs given a product design, as shown in Figure 2. The methodology uses a graph theoretical approach to integrate tolerance analysis with product, process, and production modeling. It is developed in collaboration with the Computer-aided Manufacturing Laboratory at Worcester Polytechnic Institute under support from the National Science Foundation with case studies provided by Delphi Automotive Systems. Details of the methodology can be found in a book entitled “Advanced Computer-aided Fixture Design” by Y. Rong, S. H. Huang, and Z. Hou published by Academic Press in 2005.

Figure 2. Computer-Aided Manufacturing Planning

Predictive Manufacturing Simulation

We have developed a simulation-based methodology to evaluate the dimensional and geometric accuracy of machined components, as shown in Figure 3. The methodology uses sample points to represent workpiece geometry. The change of spatial positions of these sample points are simulated based on manufacturing error synthesis. Various manufacturing errors are lumped into two categories based on their effect, namely, setup error that accounts for the deviation between ideal and actual workpiece position, and machining error that accounts for the deviation between ideal and actual cutting tool position. Virtual inspection is then conducted based on the coordinates of the final sample points to evaluate the geometric accuracy of the finished components. Details of the methodology can be found in a paper entitled “Tolerance-based Process Plan Evaluation Using Monte Carlo Simulation” by S. H. Huang, Q. Liu, and R. Musa (International Journal of Production Research, Vol. 42, No. 23, 2004, pp. 4871-4891). A software tool, VTurn, is available to demonstrate this methodology.

Figure 3. Predictive manufacturing simulation methodology.

Quantitative Aggregated Evaluation

We use multi-attribute utility analysis (MAUA) to identify the best process design. MAUA is a generic decision-making methodology. It assesses the decision-maker’s preference structure and model it mathematically with a multiple attributes utility function. It then applies this utility function to help the decision maker reach an optimal decision. An example of a simplified bearing bracket is shown in Figure 4. There are two alternative process designs (setup plans). Using predictive manufacturing simulation, we found that the first process design will achieve higher quality in perpendicularity whereas the second one will result in better precision in runout. By interacting with the product designer we created a multi-attribute utility function to make trade-offs between perpendicularity and runout requirements. It was found that the first process design is preferable to the second one. Note that this decision is influenced by both part tolerance specification and process design. A part produced using the first process design has a 100% probability of meeting perpendicularity requirement (0.02) and 98% probability of meeting runout requirement (0.03). If the second process design is used, the probabilities of meeting the perpendicularity and runout requirements are 92% and 100%, respectively. Obviosly, the first process design is better. More details can be found in a paper entitled “Multiple-Attribute Utility Analysis in Setup Plan Evaluation” by N. Xu and S. H. Huang (to be published in ASME Journal of Manufacturing Science and Engineering).

Figure 4. Multi-attribute utility analysis of two alternative process designs.

Process Optimization

Machining parameters (speeds, feed rates, and depth of cut) play a critical role in process optimization. Traditionally, companies use machinability data handbooks as a source for selection of machining parameters. The handbook values are general and represent conservative machining conditions that serve as a safe starting point. Companies must then adjust the parameters for their specific processes to achieve optimal results. This can be done using the model self-evolution functionality in our knowledge-based modeling framework. An initial general process model is used to suggest “optimal” machining parameters. The resulting process outcomes are then measured and fed to the model so it can adapt itself to minimize the difference between model prediction and the actual process outcomes. The model then suggests new “optimal” machining parameters and this process is repeated until desired process outcomes are achieved.

Process Capability Quantification

Traditional process capability analysis compares 6 sigma of a controlled process with its specification range to determine whether or not the process is capable of meeting its specifications. Indices such as C_p and C_pk are used to quantify the capability of a process. These indices are associated with a particular tolerance specification of a specific part. They can be calculated only after the part has been machined for a sufficient number of times. Such information cannot be used for predictive simulation of the machining of new parts. To overcome this problem, we focus on quantifying the root cause of process capability, i.e., manufacturing errors. These errors, mathematically represented as probability distributions, can then be used as input for predictive simulation of the machining of any parts. Preliminary methods and algorithms to estimate probability distributions of these errors can be found in a paper entitled “Simulation-Based Manufacturing Error Synthesis: Input Analysis and Validation” by R. Musa, B. C. Shultes, and S. H. Huang (Transactions of the North American Manufacturing Research Institution/SME, Vol. 32, 2004, pp. 311-318).

Metrics-based Supplier Selection

Supplier selection is a major strategic decision in supplier-based manufacturing.  Hundreds of publications pertinent to supplier selection can be found in the literature.  Business school researchers often emphasize philosophical issues and focus on developing qualitative principles to guide management decision making.  On the other hand, engineering researchers mostly treat supplier selection as an optimization problem and attempt to develop mathematical models to generate optimal solutions. We believe these two paradigms are complementary rather than competitive.  Therefore, a sensible approach to supplier selection is to develop a decision support software tool to evaluate various suppliers based on performance metrics specified by corporate managers.  We propose a set of metrics arranged hierarchically as shown in Figure 5. This set of metrics also take into account product type (i.e., make to stock, make to order, or engineer to order), supplier type (i.e., local or global), and OEM/supplier integration (i.e., no integration, operational integration, or strategic partnership) so they can be easily configured to match a firm’s supply chain strategy. We also advocate that a firm’s supply chain strategy must align with its business strategy, which is influenced by product characteristics and product life cycle stages. Details of how to determine a firm’s supply chain strategy can be found in a paper entitled “Designing Supply Chains: Towards Theory Development” by M. Vonderembse, M. Uppal, S. H., Huang, and J. P. Dismukes (to be published in International Journal of Production Economics). The configuration of supply chain metrics for supplier selection can be found in another paper entitled “Comprehensive and Configurable Metrics for Supplier Selection” by S. H. Huang and H. Keskar (submitted to International Journal of Production Economics).

Figure 5. Hierarchy of supplier selection metrics.

Technical Contact

Dr. Samuel H. Huang, Associate Professor and Director

Intelligent Systems Laboratory

Department of Mechanical, Industrial and Nuclear Engineering

University of Cincinnati

Cincinnati, OH 45221

Phone: (513)556-1154            Fax: (513)556-3390

E-mail: sam.huang@uc.edu