Virtual foundry answers 'what if' production questions; computer simulations provided Foundry Service Co. solutions to its bottlenecks onscreen - without a loss of production or labor.

The perception of virtual reality is as an advance in the video game market. But, as with many technological advances, trickle down occurs that allows for the application of these technological developments to mainstream industries. As part of this trickle down, virtual reality has entered the foundry industry in the form of simulation software.

But mention the word "simulation" to most foundrymen, and thoughts of mold filling or solidification modeling programs will come to mind. Simulation software is high-powered technology that can be used to analyze the various operations involved in making a casting. This software technology allows users to develop a working computer model of their foundry to answer the "what if" questions of alternative production scenarios and the addition, removal or adjustment of equipment and labor.

With this new technology, foundries have the ability to revamp their coreroom or molding line without a minute's loss of production or labor. If a foundry is deciding whether or not to add a furnace to its operation, for instance, simulation software can present it with the positive and negatives of the addition.

Simulation Software

The basis of simulation software is to create a representation of each step (with accurate time requirements) in a plant's material flow process. Each step is assigned a statistical distribution. For example, a core machine may always be set up with a constant cycle time for a particular corebox, so a fixed time period would be assigned. For unloading the core, cleaning it and placing it into a tote, the individual involved may have a normal time distribution with a mean and a standard deviation. The foundry model is built on the computer from the data of the distribution and deviation for each step of the process with different time elements for each core, pattern plate and casting. This enables the user to alter different parts of the production process and product mix and determine the effect on the foundry operation.

Setup times, machine failure frequencies, mean time to repair, conveyor speeds, distance between the coreroom and the molding line and scrap rates are all examples of data included in the model to increase the accuracy of a simulation. Gathering the accurate data to build a base model that truly represents foundry activities is the most time-consuming element of simulation software.

But, once completed, this base model, which is a production and process replica of the foundry, provides the foundation for alternative scenario testing and "what if" experiments. The "what if" experiments are the goal of the simulation software, as they define the potential problems or bottlenecks in the foundry system before the decision is made to hire extra labor, add or move equipment or alter processes.

'What If's' at Foundry Service Co.

After an extensive, traditional assessment of its operation, Foundry Service Co. (FSC), Biscoe, North Carolina, completed an expansion of its molding center in October 1993. In addition to adding a seventh molding line to its operation, FSC increased the power available to its furnaces, increased charges from 3000 to 4500 lb and added another bull ladle and ladle pusher. The goal with this expansion was to eliminate the bottlenecks of melt-restriction on its iron delivery system. The end result of this expansion - additional molding capacity with an overall increase in foundry production - was not to the levels that were anticipated.

At that point, another in-depth analysis of foundry production was performed using an assortment of traditional techniques - a large database and a static model - to quantify the inability of FSC to reach its maximum production output.

The model provided valuable insight into existing problems in that it pointed out delay or shortfall conditions at a given point in time. However, because it was a static model, it could not effectively address the dynamics of a jobbing operation. At any given point in time, any combination of jobs could be running on the seven molding centers. Heavily cored work, or non-cored large parts would shift metal requirements and delivery queues constantly. No one scenario would hold true. More significantly, the static model could not easily evaluate the impact of proposed changes to the operation (additional power, holding furnace, additional furnace, downstream layout changes, etc.) so, in effect, all proposed changes were pretty much a roll of the dice. FSC needed to use the accumulated data to develop a model that could address the myriad variables of its day-to-day operations, as well as effectively assess the impact of any proposed changes.

Simulation Set-up

FSC developed a computer simulation model using the Witness simulation software package, Lanner Group, Solon, Ohio. Emphasis was placed on thoroughly and accurately identifying the entire scope of the operation. The model was structured to encompass the following areas - charging, preheat, melting, tapping, iron delivery, iron pouring and molding. In addition, the model was designed to interface with an AutoCad layout of the facility to assure that all work center locations and distances reflected actual conditions.

Some basic assumptions were used in development of the model with the objective data coming from a variety of sources - including the static model, mold line production reports, furnace charge and tap reports, daily, weekly, and monthly downtime/uptime reports and daily mold line schedules.

Because of the complexity of the system, the decision was made to evaluate the model on a shift-to-shift basis rather than in a larger time scale such as a week or a month. Because FSC is a jobbing shop, the mold lines are controlled by the actual mold schedule for any particular shift. In addition to the mold schedule, iron requirements for each particular job, based on weight and yield, were also considered to more accurately model the dynamics between molding and pouring operations.

Statistical output was derived from the model by running it with schedules from different shifts. Each run was replicated 15 times to establish a 95% confidence interval. This confidence interval was used to assess the accuracy of the model. If the actual molds produced fell within the range specified by the confidence interval, then the simulation would be within 95% of what was actually produced.

After final tweaking to achieve desired accuracy, this original model was dubbed the "base model" and it was used to reflect capabilities of the existing melt and delivery system. This base model was then used for statistical comparison of data generated from secondary models that reflected the proposed expanded mold capacity, increased melt capacity and changes in process capabilities or foundry practice.

Results

The simulation produced some interesting and, in some cases, surprising data. One example was the simulation's contradiction of the earlier decision to go from a 3000 lb to a 4500 lb charge to improve melting productivity. Simulation data revealed that the increased charge size actually worked against the operation in terms of quick furnace turnaround. The simulation indicated that the normal operational pattern was to take three quick taps from a furnace, then the furnace would serve as a non-productive holder until another one or two taps were taken when the furnace could then be charged. This would lead to periods during the production shift when all furnaces were in some stage of melt cycle with no iron available to the floor. Using the simulation to do some "what if" analyses showed that a smaller charge and tighter management on size and number of taps taken per charge would allow faster turnaround and reduce the iron-related delay, even when running heavier, noncored work [ILLUSTRATION FOR FIGURE 1 OMITTED].

Some of the results and recommendations derived from the simulation were met with skepticism, so an experiment was suggested to prove out the viability of the simulation. All seven molding machines were scheduled with heavier no-core or minimally cored work and longer production runs to reduce pattern change frequency. The experiment was then run through the foundry, with the results proving the validity of the simulation, as the foundry experienced record production during the test shift.

The simulation also indicated that productivity gains could be realized through increasing the magnesium (Mg) "fade window." A hypothetical schedule of very light, heavily cored work was run through the simulation under two different scenarios, one using standard fade time parameters, the second with increased fade time. The second scenario suggested improvements, not only in terms of additional mold production, but in reduction of iron pourbacks. To prove out this theory, a second foundry experiment was scheduled, loading the molding machines with light, heavily cored work. The established parameters for Mg fade time were not changed for the experiment, but several proposed changes in tapping, treating, and slagging techniques were incorporated that effectively increased the productive portion of the fade window. Actual results during the experiment paralleled the simulation's projections.

Prior to undertaking the simulation project, the installation of a holding furnace had been considered to reduce iron delay downtime. However, the simulation data suggested that a holding furnace was not necessary to achieve significant productivity increases if several other changes were made, including the acquisition of additional power. This required the installation of new transformation equipment and a subsequent renegotiation of the electrical power rate with the utility company, but the project was justifiable based on the simulation data that was able to quantify productivity increases based on increased melt rates.

In addition, the simulation suggested productivity improvements that could be obtained by the adoption of changes in tapping, treatment, handling and delivery practices. These improvements included changes in slagging practice, elimination of cover steel, changes in the MgFeSi addition to improve recovery and the installation of signal lights to improve iron delivery and further reduce pourbacks. All of these were successfully incorporated into the operation.

The simulation projected an 8% improvement in mold production through incorporation of the changes. In reality, FSC has achieved a 22.1% improvement in mold production over the past year and a half. The simulation projected a 33% reduction in pourbacks by incorporating a different tap size and delivery routine. Pourbacks are currently running 58.4% lower than they were a year and a half ago.

These results demonstrate that if properly used, simulation modeling of foundry operations can be a powerful tool that can provide foundry management with valuable insight into current operations, assist with continuous improvement and offer the ability to answer the question, "what if?"

Cautions

Those involved with the iron delivery simulation project were very pleased with its success, but are quick to add that simulation modeling is a tool to further extend the capabilities of traditional industrial engineering techniques, not a substitute for them. The basic tools of workflow analysis, methods analysis and work measurement are essential to the acquisition of sound data needed to build the model and drive the simulation. The old saying, "garbage in-garbage out," accurately predicts the results of any simulation project that does not include sound basic data.

In addition, the product of a computer simulation is data, which must be analyzed and interpreted. A computer simulation will not "tell" you anything, rather it provides the industrial engineer skilled in statistical and comparative analyses a wealth of information from which to draw conclusions. And, it is these conclusions that lead to the solutions to problems or point out opportunities for improvement. Thus, simulation enhances, but does not replace critical thinking, common sense and good judgment.

RELATED ARTICLE: 'What if' Scenarios Tested at Grede Foundries, Inc.

In May 1996, at its Vassar, Michigan plant, Grede Foundries, Inc., wanted to increase production of its Ford molding line. This increase would be tied directly to changes in the coreroom. As a supplier to the highly competitive automotive industry, Grede invests heavily in both new foundry equipment and continuous process improvement to maintain its business growth.

"Since almost 50% of our business is with the auto industry, we have to respond to the same demands it faces relative to price and quality," said Chris Paras, corporate industrial engineer at Grede. "We have to continually improve our processes to increase production, reduce costs and increase throughput."

For Grede, this continual quest to improve, grow and increase production brought it face-to-face with simulation software.

"We've been using simulation software to evaluate our processes," said Paras. "That covers operations in work cells like the molding line and coreroom, the transport conveyors, automatic guided vehicles and some labor. We're always trying different scenarios to learn how we can operate best to meet current demands, and also how we can meet long-term goals at the lowest cost."

Using the three-dimensional, computer-driven simulation software, Quest, by Deneb Robotics, Inc., Auburn Hills, Michigan, to develop a working model of its foundry, Grede tested alternative scenarios for three different stations of the coreroom - the core dipping station, the number of workers in the coreroom and the conveyors to the molding line.

The results from the computer simulation helped determine the optimal number of workers for the coreroom, that the core dipping station should be automated, not manual, and that non-accumulating conveyors should move the cores to the molding line. The simulation also helped Grede decide to add an additional shell molding machine at the end of the conveyors to handle the increased output from the coreroom.

"We found how to increase output in the area without extra people, just by rearranging some of the operations and adding a machine. The interesting thing is, we did this project entirely on our own initiative. Since then, Ford has forecasted an increase in molds. We anticipated this demand, and used simulation to figure out how it could be met with the fewest changes and at minimal cost."

These "what if" experiments are what initially attracted Grede to this technology in September 1995.

But a simulation software program, according to Paras, isn't a tool you can load up and have running by the next afternoon. The simulation project at the Vassar plant was a 2 1/2 month process. But now that Paras has completed multiple simulations for Grede, he budgets one month from the collection of data to the testing of alternative scenarios.

Paras concludes, "Do not expect results immediately. It is very time consuming. Someone needs to be dedicated to developing the simulation and the software. But once the system is developed, it is a powerful tool for answers."

- Alfred T. Spada, assistant editor Deneb Robotics, Inc. contributed to this piece.