Case Study : Durango Paper Company
Making the most of It - Trim optimisation at Durango-Georgia Paper Company
Durango-Georgia Paper Company, Saint Marys, Georgia, has the capacity to produce 410,000 tpy of coated bleach board, uncoated free sheet, and kraft grades.
Striving to increase productivity, improve efficiency, react quicker to customer's changing needs, and increase profitability is a constant battle for today's production planners. Finding the optimal balance between these interconnected, yet sometimes conflicting goals has become part of the daily routine. In today's fast paced and ever changing world, managers are driven to pursue tools that allow them to quickly compare and evaluate alternative solutions. This effort to plan production in a manner that contributes to organisational value is what drove Durango-Georgia Paper Company (formerly Gilman Paper Company) to examine the need for an automated trim optimisation system. The mill knew that they needed to achieve greater trim efficiencies, while at the same time not compromise their flexibility to meet their customer's needs or the ability to react to various production situations.
Durango-Georgia Paper Company has the capacity to produce 410,000 tpy of coated bleach board, uncoated free sheet, and kraft grades on its three paper machines. The focus of this progressive privately owned company is to fully utilise its available resources for maximum benefit. This includes a strong focus on appropriate capital utilisation, effective employee productivity, and an unyielding focus on critical business processes. They make the most of what they have. They needed a trim optimisation system that would allow them to optimally balance the opportunities and constraints of their mill and their business.
While the production planners for Durango-Georgia are focused on a single mill site supply chain, they must deal with a number of factors that make trim planning a challenging situation. The customer base and grade structure are both very diversified. Each machine regularly runs through cycles of more than sixty grades and basis weight changes per month. This causes the demands on planning and servicing of customer orders to be just as diversified. Each machine has a committed winder with deckle trims of 193", 217", and 220"; typical customer roll widths may range from 6" up to 110". Constant demand changes from order to order force the planners to consider only a very small percentage of roll widths that may be considered 'standard' sizes, or valuable stock sizes. Flexibility to service changing customer needs is a critical competitive focus, even if the orders are very close to being produced on the machines. Limited floor space on the loading docks and warehouse is also a constraint that must be taken into consideration.
Nearly each morning throughout the year, the production schedules are updated and communicated with the production staff. The master schedules are generated manually with machine cycles covering seven to ten days with two or three grade/basis weight changes per day. Within each basis weight there may be several trim plans required due to differing diameter, core, wind, and other parameters. An average of fifteen to twenty trim plans are generated each weekday to support the master schedule. The complexity and size of each trim plan will vary depending upon many factors, but most plans were taking a scheduler from eight to twenty minutes to generate under normal conditions.
Business goals in trim planning
Planners at Durango-Georgia Paper Company are focused on meeting organisational goals. This is at the heart of finding the optimal balance between the many dynamic trim related factors that affect the overall performance of the organisation. A simple analysis of the factors that the planners must consider includes both opportunities and constraints. Some of these are; machine trim efficiency, proper allocation of production to fulfill customer orders, meeting customer delivery promises, loading dock or WIP constraints, order spread minimisation within a plan, proper stock level constraints, knife change minimisation, unique pattern minimisation, special production constraints, and value creation. While all of these factors are taken into account when evaluating trim plans, trim efficiency is usually the dominant measure of success. Each of these factors however, may be weighted more or less for any given problem depending on the perceived demands of the organisation at that particular point of time. An example is the need to balance the cost of trimming around a fault in the web due to a hole in the wire for three or four hours until enough staff are on hand to make the change, or is it better to call in the extra staff to change the wire sooner. Another example is whether or not to add lower valued stock rolls into a trim plan in order to achieve a higher efficiency. All of the trim plan impacted factors should be considered in these evaluations including overall cost versus value creation and customer demand fulfillment.
When these trim plan impacted factors are combined with the mathematical combinations of various roll sizes, diameters, core sizes, differing wind directions, and multiple machine possibilities, it creates a rather complex decision matrix. In mills with various winders, extruders, guillotines, and other secondary machines that need to be scheduled, the variability of these factors may be expanded exponentially. When this decision matrix is then applied in an ever-changing market and demand profile, it can become overwhelming. The solution that balances these factors to create the greatest organisational value would be the optimal solution.
This goes well beyond the traditional capabilities of manual trimming customer orders to fit machine constraints. The scheduler-pencil-paper-adding machine combination has become insufficient to satisfy this critical business process in today's competitive world marketplace for most mills.
The search for a system
Finding the right tool to properly fulfill the need to optimise trim plans was Durango-Georgia's (then Gilman) focus in 1997. Through a rather extensive search for suitable products on the market, eight vendors were selected for further evaluation. The factors to evaluate these systems were; cost, user friendliness, analyst capabilities, support, hardware requirements, ease of implementation, and relational ability to other existing systems. Ultimately, Greycon's X-Trim was selected as the best fit as a comprehensive trim optimisation solution.
One of the factors that impressed the scheduling staff back in 1997 was the ability to identify a dollars and cents value to the trim plans they were creating. The concept that the planners were not 'selling paper', but 'allowing the customer to have time on the machine' was not possible to adequately evaluate until this tool was implemented. This changed the philosophy of the planning process from focusing on the time consuming task of "fitting the orders to the machine" to actually thinking about the impact the plans had on the organisation. "The biggest need was to be able to quickly generate trim plans with minimal waste, low knife set ups, and that meet the customers' needs," explained Missy Brazell, Customer Service / Production Planning Assistant Manager. "Before, we just didn't have time to consider everything that needed to be considered."
System design and implementation
Greycon's trim optimisation system has been interfaced to a large variety of relational data base systems including Oracle, AS/400, MS-SQL Server, MS Access, Sybase, Informix and others, most frequently through ODBC. It is also possible to interface to systems that utilise flat file or spreadsheet input/outputs, or operate the system in a totally stand-alone mode. The system is designed for multi-user operation, so that multiple planners may work simultaneously on the same database, while ensuring that the same order does not get trimmed twice. Like many mills in the paper industry, Durango was faced with limited resources in terms of integrated information systems, the system design and implementation had be creative to meet the particular business needs of the planning staff. The trim optimisation tool needed to have near to real time information. The Durango-Georgia order entry system at the time of system design was a Cobol legacy system that is Unix based running on a HVS operating system in an Escala server. The information needed for trim optimisation was downloaded to an NT based Compaq server where it could then be used by the planning staff. This download was set to occur automatically every fifteen minutes. The tool of X-Trim itself is a pure 32-bit system that may be installed either locally on a pc or on a server. X-Trim now runs on Windows 2000 or NT platform. Durango decided to install it on individual pc's. This allows the users to access the tool from their desk pc or from a laptop in a remote location. The work to perform the interface design and implementation, beginning in October 1997, took only about three weeks, which was spread over a three-month period. In the meantime the planners were beginning to use the product manually after one day of training. After full implementation, two more days of training were conducted and the project was complete in the first quarter in 1998.
The end product
The system Durango-Georgia and Greycon jointly implemented is seen by the IS staff as a task oriented rather that information oriented system. The requirement for complex system integration was not substantial; however the power of the tool to effectively use the information it did receive was extremely powerful.
It employs state-of-the-art integer programming techniques to guarantee that answers to most problems are: optimal in terms of waste and optimal in terms of number of setups (patterns). It also employs advanced heuristics for minimizing secondary objectives such as knife changes and order spread.
The planners at Durango have the ability to visualise the trim plans based on roll width, order number, customer names, and other information associated with each roll. They can also graphically edit the solution on this screen.
Graphic Editor within Greycon's X-Trim allows the solution to be viewed by multiple criteria and then edited on screen. Trim efficiencies at Durango-Georgia were immediately impacted as planners focused on making the most of the machine deckles.
Paper machine faults and position requirements
The tool can deal with order-specific position constraints (e.g. "middle", "not left"), but also it can cope with multiple imperfections on the web and deckle around them. "It can easily save a couple hours of work when I get calls in the middle of the night to retrim around defects," said Darby Banks, Production Planner.
Multiple primary machine options
Durango needed a system that could cope with multiple paper machines when one order or grade may be produced on more than one paper machines. The program determines how best to split the problem according to optimisation and run length constraints. This has been used by Durango to determine which machine a grade or group of orders will be most optimally run on. Additional costs may be specified to take into account if there is extra cost of transporting from a distant site.
The planners also frequently use a feature called optional rolls. This allows the planner to input a roll that may be used to improve the trim efficiency, but is not required to be manufactured in the current order group. The planner is able to minimize the amount of future order or inventory generation. Paul Blackmon, Senior Vice President of Sales and Marketing said, "Our focus on inventory control is due to the negative impacts that it has on cash flow, as well as the fact that inventory obsolescence is occurring at an ever quickening pace. Inventory to accommodate JIT is not acceptable, we must focus our business processes on satisfying customer demands more efficiently."
Knife change and set up minimisation
The system employs innovative algorithms to minimise the amount of knife changes and the number of unique patterns. These techniques, some of which are based on genetic algorithms, work without affecting the waste of the solution. Durango was able to minimise setups by over twenty percent with this application.
Reports in X-Trim are based on Crystal Reports and can therefore be customised completely by the end-user or local IS staff. Reports can include graphics, bar-codes, bitmaps and may be exported in a variety of formats, including HTML, Lotus 1-2-3, Excel, Word and ODBC formats. At Durango, the planners are able to provide a detailed report to the machine personnel that includes information about customer preferences, order parameters, patterns of the solution, as well as overall measures of the plan, such as waste as a percentage or in pounds. Other information available through reporting is a summary of the rolls widths and core sizes contained within the solution, which is used by the in-house core cutters.
Planners may evaluate different solutions graphically based on multiple Criteria. These criteria may be selected by the planner.
The organisational impact
The impact of implementing X-Trim was immediate. The schedulers upon realising the powerfulness of the tool and the potential impact it could bring, began using the system by manually, keying in customer demand data until the IS staff was ready to implement an interfaced import. In doing so, the project nearly paid for itself before it was fully implemented. Trim efficiencies were improved by nearly half a percent.
This allowed the mill to gain four and a half tons per day at no additional production cost. This accounted for more than 1600 tons by the end of the first year. A scheduling office often has the reputation of a hectic place, but because it was possible for these complex mathematical problems to be solved so quickly, the planners now actually had time to consider the impact of the solution and then rerun the problem as many as needed to achieve the desired result. The planners were also able to graphically compare multiple solutions to see how each had an effect on selected trim related factors, such as set ups, waste generation, and order fulfillment. See Figure 4. In addition to these successes, there were also some problems that arose after implementation. One mistake early on was made in the push to achieve higher and higher efficiencies. The planners would occasionally over burden the loading areas and shipping crews by combining too many customer orders together to achieve greater mathematical possibilities. The schedulers were also able to achieve such an efficient trim on a consistent basis that the machine staff had to reevaluate the defined maximum paper machine trims for various grades. The trim plans were actually pushing the paper machine operators to look for ways to become more uniformly consistent across the web. Both of these situations were resolved through some heartfelt conversations between the various staffs. The improved trim plans caused the need for the business processes to be defined, evaluated, and communicated.
Because the implemented product was able to more effectively and quickly allow the generation of acceptable trim plans, the users felt less pressure to create the plans 'early'. They were able to wait closer to the time of production to generate the trim plans. This was beneficial for several reasons, 1) the scheduler typically had a larger group of data to attempt to achieve a higher efficiency, 2) the amount of rework was reduced that was caused by changing demands after a plan was in place, and 3) customer requested order changes were deemed more acceptable to the process so there was less resistance to accepting the changes to meet the customer's needs. There have even been cases when a newly optimised trim plan is being produced on the shop floor within the hour of receiving the order from the customer. This was an unexpected advantage of 'getting closer to the customer.'
Overall the project was deemed successful and after only three months of using the product, full pay back occurred. The system has been in place for four years now and like a hammer to a skilled carpenter, has become an integral part of adding value to the organisation by the planner's daily use. The task that required a staff of five in 1997, now only requires two persons. Durango-Georgia staying true to their philosophy of making the most of their resources is now in the process of applying this technology at the shop floor level. The winder crews will be able to use this optimisation technology for retrim when quality defects or other unforeseen conditions prevent the original plan from being produced.
What is a mill to do to ensure that production is being planned properly? First, the management must provide proper training so that the planner understands the complexity and the outcome of the decision making process. Second, they must also be absolutely certain that the planner understands the goals of the organisation. This is sometimes an overlooked need because it is assumed that the planner is on board with the organisational goals due to his/her years of experience. There was a quote from a past PIMA conference that went something like this, "The three most powerful people in the world are, the president of the United States, the premier of the Soviet Union, and the captain of a nuclear submarine:they forgot one: the guy who schedules your paper machine." Last, the planner must be provided a user-friendly tool that will quickly and easily assist him/her in this often-complex decision-making process. This tool must also provide them with the visibility to analyse the impact of those decisions towards creating organisational value.