01 April 2004
General to master
Capturing best engineering practices in recipe generation.
By Dennis Brandl
General and site recipes are the least known elements the ISA-88 ISA Batch Systems Standards describe. However, they can provide significant economic benefits to companies that manufacture the same products at multiple sites or use contract manufacturing. They may also speed up new product deployment and aid in business-to-manufacturing integration.
One of the most controversial aspects of general and site recipes is the process of transforming a general or site recipe to a master recipe. Companies using general or site recipes have traditionally viewed generating master recipes as a manual engineering task. The transformation has generally required significant experience in master recipe construction and detailed knowledge of the target process cells. Historically you needed an understanding of the economic or operational optimizations important to the plant and an understanding of the notations and terminology of the development or pilot plant.
The ANSI/ISA-88.00.03-2003 Batch Control Part 3: General and Site Recipe Models and Representation standard contains a description of transformation that gives a basis for automated transformation of general recipes to master recipes. This section had significant discussion and disagreements, because the algorithms for transformation are not well known and can be hard to implement. However, you can automate the transformation if the general recipe definition strictly conforms to the ISA-88 Part 3 model. Even if transformation is not automated, following the ISA-88 Part 3 standard simplifies a difficult engineering task, resulting in more consistent master recipes, reduced test batches, and faster master recipe generation.
The above methods have seen use in multiple companies. These methods allowed the companies to gain economic advantages from general recipes, reducing the engineering time and cost for new product deployment. General recipe transformation can use each site's best engineering practices in the transformation process. The result is more consistent master recipes, fewer trial batches, and improved multisite product consistency. You can manually perform methods for general recipe transformation, or you can use semiautomated or completely automated methods. Automation requires strict conformance to the ISA-88 Part 3 procedure definitions models.
General and site recipes
The information manufacturers need to make a product is known as general and site recipes. This information defines the basic chemistry and physics a manufacturer would need to make the product without defining specific equipment, unit layout, unit transfers, unit initialization, and unit cleanup. You could view general and site recipes as material-centric, rather than equipment-centric (master) recipes. They may contain constraints on the target equipment, usually when the equipment characteristics can affect the chemistry or the physics of the desired reactions. Typical constraints are the interior lining or material of vessels and pipe construction, mixing types (normal or low shear), and heating ranges or rates.
Where to begin
The first step in effective transformation is for each site to define the best way to implement each process action on each unit that can perform the action. The ANSI/ISA-88.01-1995 Batch Control Part 1: Models and Terminology recipe standard is a good way to do that. There may be a simple one-to-one relationship between process actions and master recipe phases, or there can be a very complex relationship. In ISA-88 Part 1, a new structure is a segment of a master recipe. It is called the master recipe transformation component. The standard defines it as "the part of a master recipe that is used in the transformation of an equipment independent recipe into a complete master recipe." Some users of general recipes use the terms recipe segment, recipe component, or (as in this article) transform component in place of the long name. A transform component may be as simple as a single phase or as complicated as a set of complete unit procedures.
Each site should define the appropriate transform component for each unit or process cell that can perform the action. Transformation of a general recipe to a master recipe uses the mappings, in part to determine which units are possible targets to the master recipe. Actual use of transform components in transformation must also take into account the range of values for parameters in a process action and the corresponding range of parameters within the phases of each transformation component. For example, a process action "HEAT" may define a temperature of 220°C in a general recipe. When determining possible target units, you will need to check the 220°C value against the actual range of the heating phases to ensure the required temperature is within the range of values available for the unit. Not only must a unit be able to implement a process action, but it must be able to meet the requirements contained within the process action parameters.
Other elements of information you need to define are how to prepare each unit before using it and how to clean up or finish up each unit after using it. Initialization activities may define such operations as getting the unit up to temperature or pressure and making sure the unit is clean or empty. Finalization activities may define such operations as cleaning, sterilizing, flushing, and cooling the unit. Again, define the initialization and finalization operations as master recipe transformation components. You do not need an initialization or finalization transform component for each unit. Storage units may have no initialization of finalization operations.
Once all basic information is in place, transformation can start. First, transform components for each process action available on each unit or unit class. Transform components for material entry to a unit or unit class, for material exit from a unit or unit class, and for material transfer between units. Then you will need to transform components for unit or unit class initialization and finalization.
The first transformation step is to expand the general recipe using the underlying process. If the places where materials are joined is used as a boundary condition, then we can call each collection of process actions a material path because it defines a path of actions for a single material dependency. The next step is to eliminate transform components on units that do not meet the equipment requirements associated with the process stage or process operation. A process stage may require equipment with external heating coils, because the material would foul internal coils. If the equipment (unit) associated with a transform component has internal coils, you would remove that specific mapping from the list of possible mappings.
If you cannot find any transform components on units that meet the equipment requirements, then no transformation is possible. Also, effective use of equipment requirements can simplify the transformation process—if you eliminate inappropriate equipment early.
All possible combinations
Using the remaining transform component actions, build a list of all possible combinations of transform components and units that can perform the required set of actions. You could call these transform paths because they define possible transformation paths through units.
You can identify each transform path (TP) by its starting unit and ending unit. The right side of the figure below illustrates the combination of the transform paths with the material path. In the example below, there are four possible ways to perform the right top material path segment and two possible ways to perform the final material path segment. At this point it is sometimes possible to eliminate transform paths because of illegal transfers between units. For example, if a material path required a transfer from Unit 1 to Unit 3, then the associated transform path could be eliminated because there is no legal transfer defined between these units. In the examples there are multiple possible transform paths for a single material path; in practice there is usually less complexity because many process actions are limited to specific units. In the figure below, the identification of a transform path indicates the starting and ending unit, for example (U2-U5) indicates the transform path starts in Unit U2 and finishes in Unit U5.
Next you need to construct each possible combination of transform path and material path. Each combination can be called a recipe path. Each recipe path is a possible combination of transform components and material dependencies that can perform the actions the general recipe requires. Obviously this can generate a large number of possible recipes. Eliminating any path with an illegal material transfer can reduce the number of recipe paths. If any recipe path required a material transfer from Unit 3 to Unit 5, you could eliminate it.
Hopefully, you will have at least one recipe path remaining after considering undefined unit-to-unit transfers and equipment requirements. If no recipe path remains, then you will not be able to construct a valid master from the general recipe. If that is the case, you will need additional unit-to-unit transfers or process action-to-unit mappings. If multiple recipe paths exist, you can use various optimization strategies to determine the best master recipe. Or you can generate multiple recipes, and the schedule can define which recipe to use. For example, there may be a cost or efficiency associated with each unit or with each process action on a unit. You can calculate the cost or efficiency for each recipe path and choose for transformation the path with the lowest cost or highest efficiency.
Transform paths (left) and material paths (right)
Construct the elements
The next steps in transformation involve generating the master recipe from the recipe path. There are two main elements of master recipe generation: building the operations and building the unit procedures. The general steps are:
- For each unit used, create a unit procedure.
- When you first use a unit, add the transform component associated with unit initialization. A good practice is to make this the first operation in the unit procedure.
- When you last use a unit, after material is transferred out, add the transform component associated with unit finalization. A good practice is to make this the last operation in the unit procedure.
- Add the transform components associated with each process action.
- Use the process operation boundaries to define operation boundaries.
- When process actions are in parallel, create a parallel structure within the operation.
- For each material transfer between units, add the transform component for each unit into each unit's operation. This step actually defines the overall structure of the master recipe.
Material dependency map and process action to unit mapping
Master recipe structure
Because you organize master recipes by unit procedure, it is vital that you determine the correct unit procedures. The material transfer between units actually defines the overall master recipe structure. Apply the following rules to master recipe structure:
- For each starting unit in the recipe path, add a unit procedure under a common parallel. Eventually, there may be timing phases or allocation elements added before these unit procedures, but theoretically they could all start in parallel. Practically, there are usually timing considerations you should consider. Path P6 may take six hours, while Path P5 may take twenty minutes.
- If a transfer occurs and no parallel process actions are required during the transfer, then you can place the transfers in separate unit procedures from the previous processing.
- If parallel actions are required during a transfer out and the transfer in, then you will need to define the unit procedures in parallel.
- If there are parallel actions on the transfer out but not on the transfer in, then the transfer in can be a separate parallel unit procedure, but the transfer out cannot.
- If there are parallel actions on the transfer in but not the transfer out, then the transfer out can be in a separate parallel unit procedure.
- You might need to add deallocation elements after the unit finalization.
- Some transformation components may involve multiple units, such as preparation of a material mixture. Where possible, implement these as separate unit procedures, using the same rules above for material transfer rules.
The final step in master recipe generation involves mapping the parameters in the process actions into phase, operation, and unit procedure parameters in the master recipe. There may be a one-to-one correspondence between the parameters, but it is more likely there is a complex relationship. For example, a process action parameter may have a single value for the amount of a material addition, and you could implement this using several phases. Phase 1 could add 25% of the total amount using a shot add method. Phase 2 could add 50% of the total using a rate controlled add method. And Phase 3 could add the remaining 25% using a temperature controlled add method. You should copy formula values and header information from the general recipe to the master recipe, along with any equipment requirements beyond those you use in transformation.
Behind the byline
Dennis Brandl is president of BR&L Consulting in Cary, N.C. He presented at the World Batch Forum North American Conference in Woodcliff Lake, N.J., in April 2003.
One link at a time: A supply chain success story
By Jane B. Lee
Optimizing the supply chain is a hot issue in the race for global competitiveness. Defining the perfect end state is easy; getting there while you have a business to run is not. To achieve dramatic and sustainable results, you'll need to put in place a process for improvement, not a one-time project.
Between 1991 and 1995, our DuPont business did just that by tackling one real-world problem at a time. We combined substantive business-process changes with key information through cross-functional and interregional teamwork and support systems to cure visible deficiencies. We needed to agree on a forecast beyond annual profit objective. We needed to establish clear responsibility and accountability for forecasts and inventories and consistent, high-quality information flows among regions. We also wanted to foresee potential service problems before customers were affected.
The resulting supply chain management process reduced total working capital by 25% and improved customer service and schedule stability. Both process and systems are dynamic enough to change as rapidly as business emphasis and market conditions dictate.
Where we began
Our business has been in the forefront of supply chain management initiatives for more than ten years. In the midst of constrained capacity, our special polymers business operated on commodity-sized units. We recognized the need for visible integrated demand, inventory, and production data. But we didn't necessarily do everything right the first time on the path to supply chain optimization.
Fundamentals of good supply chain management were already in place in 1991 when the business brought in outside consultants to lead us through a thorough analysis of the U.S. supply chain forecasting. We discovered small customers placing same-day and next-day orders were getting better service levels than steady, larger customers who gave orders with long lead times. Customer service averaged 87.5%. And the biggest reason for failure was product unavailability.
It took a crisis involving Europe to kick start a real step change in supply chain effectiveness—and to make us recognize the critical nature of a global focus. With inventories high and pressures mounting to meet regional targets by year-end, Europe began canceling orders on the U.S. We moved quickly to consolidate European warehouses and order entry and inventory systems. Within four months we created a model tied to European and U.S. order-entry systems and to the freight forwarder system. European PCs now had daily visibility not only of on-hand inventories but time-phased forecasts, orders, and restocks—including in-transit material by estimated arrival date. The system not only calculated inventory standards based on optimum order cycles and restock quantities, but also generated recommendations for adding or delaying restocks to keep inventories in balance. PCs in Europe now had the tools they needed to manage inventories and restocks effectively.
Supply crisis sparks breakthrough
By now it was early 1994, and demand suddenly went through the roof. At the same time, several unusual manufacturing upsets curtailed our supply. Suddenly, it became clear the real question was, "How can we best assure product availability for all our customers all the time?" Faced with an undeniable need, we could abandon the old forecasting system and have access to twelve months of historical data at the line-item detail level (product, package, customer, ship date), along with month-to-date sales and all open orders.
Customer service representatives were already using the renewed available-to-promise model. We put new policies and procedures into place to ensure out-of-pattern demand triggered an immediate call to product coordination. Online reports enabled PCs to determine whether such an order might jeopardize regular customers who just hadn't placed their orders yet this month and to incorporate the new demand pattern into production planning as necessary. Concurrently, our consultants built structures to record the regional 1+6 forecasts and online reports to let different customer service representatives handling specific countries see whether a requested order was still within the expected regional total. As with domestic orders, out-of-pattern demand triggered a call to product coordination.
We had hit the mother lode—a process and supporting system that would provide not only a meaningful forecast but also continuous feedback loops for demand versus expectations. The immediately implemented functionality got us through the supply crisis with minimal customer impact. Product coordinators with twenty years of experience said they had never seen supply so tight. Over the next several months, we added refinements. Demand monitoring brought to light not only increases in demand but also those product and package substitutions that necessitated realignments but that we had never systematically captured before.
Our ability to see demand changes quickly allowed us to improve what customers value most—reliability of promised delivery dates—while keeping inventories low. Service levels have improved from 87.5% to 97%. Schedule change and expedited transportation costs became so low, reports were deemed non-value-adding and eliminated. The biggest benefit, though, was the ability to foresee and prevent problems before they actually arose. ?
Behind the byline
Jane B. Lee is a senior consultant at Supply Chain Consultants in Wilmington, Del.