Integrating Manufacturing Into the Supply Chain Model

Supply chain management is a topic that has generated considerable interest from businesses in recent years. This interest is understandable given the magnitude of the benefits that can be achieved when business entities begin to comprehend and manage the entire chain of processes required to convert market demand into profit. This perspective of global optimization reveals that the profitability of a business is dictated not only by how efficiently the business executes its own functions, but also how well it collaborates with all the other businesses that form the supply chain — or, more accurately, the value chain. This value chain consists of all the resources and processes required to generate and quantify the demand for a product, acquire materials, turn those materials into product, deliver the product to customers, and collect revenues from those sales. This is described as a chain because each process is dependent on the rest of the chain, and a failure in any one process affects all the businesses in the chain.

The goal of value chain management, then, is to maximize the profit derived from these interrelated processes by accelerating the velocity of transactions up and down the value chain, thus reducing the amount of money tied up in interim process steps and improving customer service. Because a business traditionally will invest in buffer inventory to protect itself from unreliable supply, greater visibility and communication of demand and capacity signals improves reliability and decreases the required buffer.

The theory of constraints addresses these issues and uses clear definitions of terms and the principals of factory physics to put a structure and logic around how to accomplish the ultimate goal of any system of dependent processes. To illustrate how the theory of constraints aligns with the goals of value chain management, let's briefly summarize its concepts.

A system is composed of a group of related and dependent processes (Fig. 1). The throughput of the system, which is defined as the rate at which the system generates money, is limited by the most constrained resource, called a bottleneck. Therefore, all resources in the system should work to ensure that the bottleneck never loses productive time by providing a constant supply of inventory for it to process. This cooperative effort by all of the system's dependent resources is referred to as seeking global optimization (vs. local optimums). The goal of the theory of constraints is to maximize system throughput while maintaining the minimum amount of inventory and operational expense. The minimum required inventory is maintained by only feeding inventory into the system at the rate that the bottleneck can process it. This, in turn, requires that we have an accurate assessment of the capacity of the bottleneck at all times. Another benefit of the reduced inventory and increased throughput is that the cycle time for the system is minimized.

 
1. The goal of the theory of constraints is to maximize the throughput of the system while maintaining the minimum amount of inventory and operational expense. The minimum required inventory is maintained by only feeding inventory into the system at the rate that the bottleneck can process it.

In the realm of manufacturing, the theory of constraints has been widely adopted and has consistently shown impressive results when effectively reduced to application. One of the most successful examples of this reduction to application uses the AutoSimulations product, Real Time Dispatcher, with the focus being on enabling all toolsets in a manufacturing system to dispatch material in such a way as to provide a consistent feed of inventory to the bottleneck. This approach reduces idle time on the bottleneck and therefore increases the capacity of both the bottleneck and the manufacturing system as a whole. The next step is to regulate the orders into the factory to a rate consistent with the factory's capacity. Clearly, accepting and planning manufacturing orders is a value chain issue, which demonstrates the need to expand the application of the theory to the full value chain.

Some current supply chain models are based on the theory of constraints, and these models have been successfully implemented in many businesses. However, in industries with complex manufacturing processes, there have been significant problems with integrating the manufacturing process into the supply chain model. There are a couple of reasons for this. The first difficulty involves the ability to predict capacity and the second is an issue of integrating pull systems with push systems.

We have stated that we can reduce cycle time and inventory costs by limiting the input of material into the system based on the rate that the bottleneck can process it. Therefore, it is critical that we understand the capacity of the bottleneck at all times. The longer and more complex the supply chain is, and the longer and more complex the manufacturing process is, the more difficult it is to assess and predict capacity. In a single manufacturing facility the capacity analysis can be provided by discrete event simulation. But, in a complex supply chain with multiple manufacturing facilities and multiple sources of demand, the level of detail required for this type of model is not practical.

Simulation modeling is the best tool available to analyze capacity in a predictive way. However, it requires a great deal of detailed information, it takes a while to get results and, most importantly, it is not deterministic. The results must be analyzed, interpreted and iterated to reach a desired result. This makes it very difficult to integrate simulation modeling into an automated large-scale planning system.

On the other hand, no other capacity analysis tool has been found that can reliably predict capacity availability far enough in advance to facilitate order acceptance. Therefore, value chain models must accept orders based on limited capacity information that may overestimate or underestimate actual capacity. Acknowledging these limitations, we must find a practical method to control inventory.

Businesses that are pursuing advanced planning systems have generally subscribed to the theory of constraints as the basis for the value chain model. These theory of constraints-based models are considered "pull systems" by the fact that they pull material into the system based on its capacity. To accept an order, the model assesses the availability of capacity on a constrained resource at a future point in time when the order will need that resource. However, once an order is accepted into the system, it essentially becomes a push order with a committed ship date. So as far as the manufacturing facility is concerned, the starts plan is a push system.

It follows that the same companies implementing this type of value chain model would also implement manufacturing systems using the theory of constraints. As stated before, Real Time Dispatcher allows for the implementation of very powerful manufacturing solutions designed to consistently and linearly feed the bottleneck resource, resulting in lower cycle times and higher throughput. These encouraging results often entice management to make more aggressive capacity commitments. The danger here is that the higher starts commitments may exceed current capacity and drive inventory up, thus negating the efficiencies they seek to take advantage of. To realize and maintain the benefits that a theory of constraints solution can offer, the manufacturing facility clearly needs the ability to regulate the flow of material into the manufacturing system.

It is easy to see that the desire to implement these principals at the value chain level and the manufacturing level have come into conflict. And yet the benefits to be realized in either application cannot be ignored, so how do we resolve the problem? The resolution is to allow nested systems.

Another principal of the theory of constraints is that, once material enters the system, it free-flows to the bottleneck, where it enters the queue and forms a time buffer for the critical resource (Fig. 2). By observing the size and rate of growth of the bottleneck queue, we can gauge the consumption rate of the bottleneck resource and therefore the throughput of the entire system. The continual change in the size of the queue/buffer is considered normal and is not a concern unless the buffer begins to grow out of control, or becomes too low to support uninterrupted feed to the resource.

2. By observing the size and rate of growth of the bottleneck queue, it’s possible to gauge the consumption rate of the bottleneck resource and therefore the throughput of the entire system.

The approach discussed here is designed with certain limitations of current technology in mind. As stated before, a major challenge in implementing value chain management strategies is the ability to accurately predict the capacity of a resource for a future time slot. This predictive ability is currently only practical through discrete event simulation, a method that is relatively slow and, more importantly, is not inherently deterministic. A further complication is that predictive models are designed to start with a set of assumptions and then extrapolate forward for expected future results. When simulating a manufacturing facility the starts plan is an input (or an assumption), but for the purposes of value chain management the optimum starts plan is a desired output. In other words, we have a "chicken or egg" problem where the planning function wants to know what capacity manufacturing has available to determine the final product mix, and manufacturing wants to know from planning what the product mix is to state capacity. A final complication is that it is highly desirable to automate the entire solution to give real-time responses to customers for order commitments. There have been attempts by various companies to integrate the accuracy and predictive capability of simulation into the supply chain solution, but the author is not aware of any truly successful implementations.

The vision for a next-generation value chain management model that integrates discrete event simulation would not only require that the individual facility models be able to automatically verify statistically valid, but they must also be able to accept and execute on a proposed starts plan. The facility model would then pass the results back to a master model that would evaluate the results of all manufacturing facilities being scheduled. The master model could then modify the new inputs in a way that would allow the throughput of the entire network of facilities to meet or exceed customer demands. The modified starts plans would then be passed back to the facility models for validation. This process would be iterated until a satisfactory capacity allocation was reached.

The integration, automation and optimization issues, combined with the long execution times required to use simulation models in this way, pose a formidable task. The use of nested theory of constraints systems represents a more attainable and rapidly implementable approach to integrating advanced manufacturing systems into the value chain model.