Issue: Volume: 23 Issue: 2 (Feb 2000)

SimFactory2000



by Caren D. Potter

When solid modeling was first introduced, the hype was incredible. Solid modeling was going to change the world for manufacturing companies by providing a better way to define products than with stacks of 2D drawings. Lately, we're hearing the same buzz about factory-simulation software. These tools are supposed to be the next must-have programs for manufacturers because they make it possible to optimize virtual production processes before setting up a physical plant.

Some of the excitement is warranted. While the technology underlying most factory-simulation software isn't new, the tools themselves have been enhanced with 3D graphics and nicer interfaces, making them more widely accessible and applicable. But perhaps most of the excitement stems from the fact that this category of software is likely to drive the future growth of CAD/CAM. According to the market-research firm D.H. Brown Associates, the factory-simulation market is now at the point where solid modeling was 15 years ago, and it has the potential to be bigger than the product-design market.

It's helpful, during times of high-tech hyperbole, to do a reality check to see if the rewards of implementing the new technology truly outweigh the risks. In fact, back when solid modeling was being touted as the design equivalent of sliced bread, it was not uncommon to hear actual users utter discouraging words such as, "We don't use it; takes too long." With that in mind, we asked some of the leading users of factory-simulation software to discuss how their companies are employing these virtual tools and what real-world benefits they are achieving.

Gene Coffman, a staff technical specialist at Ford Motor Company, started working for the Dearborn-based automaker in the early 1980s, doing essentially what he does today-simulating manufacturing operations. Thus, he brings nearly a 20-year perspective to the discussion of factory-simulation software. "When I started with Ford, we did manufacturing-system simulation on a mainframe with little or no animated graphics, so the progress has been tremendous," he says. "Today's tools are easier to use and, thanks to hardware advances, the time it takes to run a simulation has dropped significantly."

Because each program has specific strengths, Coffman says, Ford uses a huge variety of factory-simulation tools. Indeed, the automaker has implemented a suite of tools from Engineering Animation Inc. (EAI), including VisFactory for process visualization, FactoryCAD for plant layout, FactoryFlow for material-flow analysis, and Jack for ergonomics simulations. The company also uses RobCAD from Tecnomatix for workcell simulations and a combination of Witness from Lanner Group Ltd., Simul8 from Simul8 Inc., AutoMod from AutoSimulations Inc., and Arena from Systems Modeling Corp. for discrete-event simulation.
After performing discrete-event simulations to optimize processes, Ford uses EAI's FactoryCAD (above) to lay out plant operations.




Rather than try to simulate all manufacturing processes, Ford focuses mainly on operations that were deemed critical for past programs. Engineers must first prove the business case and set objectives for any given simulation before they begin developing, testing, and implementing a model, says Coffman. "Determining the question or questions you want the simulation to answer is critical to success because models developed for one purpose can rarely be used to address a different objective or question." It also determines the type of software used and the level of detail that needs to be modeled.

A simulation of the manufacturing processes at Ford's powertrain plant in Romeo, Michigan, is typical of factory models created at the company. Producing a powertrain involves a series of machining operations. Each machine experiences a different amount of downtime due to differing levels of tool breakage and varying tool-replacement schedules. One goal of the simulation was to determine how to "decouple" the individual stations, so each could act independently and thus generate the maximum possible throughput for the overall process. In other words, Ford engineers used the simulation to make sure that if one machine went down, there would be an optimal amount of back-up inventory to keep the line moving without tying up too much material and storage space.

The results of this discrete-event simulation, which is performed by programs such as Witness, might tell the manufacturing planner that, optimally, five units of inventory would be required between station one and station two. At this point, a factory-layout program such as EAI's FactoryCAD would help the planner determine whether there's enough space for a conveyor to accommodate the extra units or how many storage racks are necessary and where to place them. "Discrete-event simulation is typically done first, independent of the layout," Coffman explains. "Then, we draw the physical dimensions with a factory-layout package to make sure we get a footprint that fits."
Ford created this flythrough of a tire-mounting operation on an SUV assembly line-using EAI's FactoryCAD and VisMockup software-to evaluate the automaker's material-handling system.




One of the recent improvements in factory simulation, according to Coffman, is that CAD data can now be imported into simulation models. For example, if the goal is to determine whether a robotic welder can reach all the welds on a new vehicle frame, the CAD model of the frame can be imported into the simulation program (RobCAD, in this example) rather than re-created. The simulation model also requires detailed information about the robot's motion as well as specifications about the controller that runs the robot. Sometimes that information is available from the tooling manufacturer. Other times, Coffman and his counterparts must collect this data from the factory floor.

In Coffman's opinion, the main limitation of today's factory simulation software is that, for the most part, the different kinds of simulation tools-discrete-event simulation, material-flow simulation, workcell simulation-operate independently of each other. He acknowledges vendors' efforts to integrate them, such as EAI's Open Virtual Factory, as well as some work EAI does specifically for Ford. But in his opinion, this is where future development should focus.

"One of the fundamental rules of manufacturing-process planning is that sometimes you want to 'sub-optimize' components," Coffman says. "In other words, individual components might have to operate at less than optimal capacity for the good of the overall operation. For example, what might be best from a materials-handling standpoint would be to arrange workstations to minimize total travel distance. But this might not be the best arrangement in terms of the process. If I'm using one tool to optimize a manufacturing process and another to optimize material flow, I can't find the best trade-off. The ability to have different simulation models interact will give us even greater benefit."
On the virtual SUV assembly line, simulations of digital workers-created with EAI's VisJack-helped Ford estimate human work loads.




Determining the return on an investment in factory-simulation software is a difficult task. The primary reason is that the money you end up saving is usually money you didn't plan on spending in the first place. "Simulation, when applied before a system is installed, will find problems that were not anticipated, preventing costly launch delays and cost overruns," Coffman says. "Similarly, when simulation is used to study improvements to existing systems, it will tell you where to avoid spending money on changes that do not generate the expected improvements, or it will verify that a change will generate the desired benefits. I believe this impact is far greater than the dollar amounts that we can identify."

Where Ford does identify actual dollar savings, it associates them with either increasing the capacity of the system or reducing investment. The former stems from identifying and alleviating bottlenecks in the process, while the latter results from identifying unnecessary expenditures or less-expensive alternatives that provide equal or better overall system performance.

For example, in a recent simulation study involving three different types of pallets used to carry material through the assembly process, the company found that the original plan called for too many of two types of pallets and not enough of the third. Using the model, they were able to determine how many of each type was needed. The same model also identified a section of conveyer that was not needed because parts could be routed through other paths without affecting throughput.

"I typically use the round figure of a half million dollars to indicate the minimum value of applying simulation to a manufacturing process," Coffman says. "This is a conservative number based on a wide range of simulation applications." In fact, the Ford-owned enterprise Visteon Automotive Systems was recently honored for its use of factory-simulation software to increase the production of parts for Ford and Lincoln sport-utility vehicles and Ford 4x4 trucks. The result of the simulation was a productivity improvement of more than 30%-an "extra" 145,000 front axles produced between January 1997 and July 1998-and a cost avoidance of $10 million for expansion of the line.
BMW employs Tecnomatix Digital Factory simulation software to plan and optimize spot-welding and manual operations at the company's plant in Munich, Germany.




In mid-1997, BMW implemented a digital manufacturing strategy partially in response to the 50% cycle-time reduction the company has achieved. Indeed, one consequence of the shortened vehicle-development cycle is that the company's manufacturing planners must now do their jobs with less information than they had in the past. "The planners used to have a complete bill of materials and a complete product definition," says Joachim Weismueller, general manager of process information technologies at BMW. "Now they have to start their job before that input is available. Our mission is to give them tools to logically plan the manufacturing process from a rough part description."

The tools to which Weismueller refers are Tecnomatix's RobCAD, Simple++, and Process Planner for factory simulation and Bentley Microstation for factory layout. And the philosophy to which he adheres is that it is not enough just to have these tools; there must also be a systematic approach to ap plying them.

"If we identify risks from the product side, such as a new cockpit or assembly sequence, we will simulate the manufacturing process during the product-development phase," Weismueller explains. "Manufacturing planners may identify additional risks from their point of view, such as potential sequence or supply problems, and these are simulated as well." But BMW does not need to simulate every part of a vehicle's assembly. Many things can be figured out just with BMW's CAD system, Catia from Dassault Systemes, he says. "For instance, in simple cases, the process planner can understand the feasibility of an assembly sequence by just seeing the parts in their locations with a 3D animation."

Like Coffman at Ford, Weismueller believes there is a lot to gain from integrating the various programs used for simulating factory operations. In fact, integration has been the focus of much of BMW's recent efforts in this area. For instance, BMW worked with Tecnomatix for the past year and a half to develop Process Planner, a program that Tecnomatix released for sale to other companies last fall. This program links product data with manufacturing operations and resources, and creates a process model that can be reused on future projects. "Process Planner is the method by which product design communicates with manufacturing logic," says Weismueller. "It lets us answer questions such as, 'Will this product fit on this assembly line?' These are questions we formerly answered with hardware."

The company is currently working on integrating RobCAD and Microstation into the Process Planner environment. It is also working to link Microstation and Simple++ so that an existing layout model can be imported into a discrete-event simulation.

Although BMW won't disclose actual return-on-investment figures for factory simulation, Weismueller says, "We are certain that the digital manufacturing project will pay back its cost with the first project that makes full productive use of this methodology."

As a supplier of limited-quantity items such as the Space Shuttle, Boeing Reusable Space Systems's (RSS) manufacturing operations differ from the mass-production systems of Ford and BMW. According to John Caddick, manager for advanced manufacturing technology at Boeing RSS, factory-simulation software plays two key roles in his organization. "We use these tools during the product-design cycle to ensure that designs are highly manufacturable," he says, "and we use them downstream to optimize our manufacturing processes."

Optimal manufacturing is vital for Boeing RSS because the company doesn't have the luxury of running a few parts through the process to work out the bugs. "When we build the product one time, it has to be like we've built it 100 times already," Caddick says.

Boeing engineers use Deneb/Assembly for what they call "digital preassembly," that is, assembling parts in software prior to the actual procedure. They also use Deneb's Quest software for discrete-event simulation and Dassault Systemes's CCPlant for factory layout. These packages are well integrated with each other and with Catia, according to Caddick. For example, Catia CAD models of both products and tooling can be imported into both CCPlant and Deneb/Assembly. Moreover, CCPlant and Quest are linked so that when manufacturing planners run a discrete-event simulation, they can watch it take place within a digital replica of the actual factory.

Boeing RSS has found that using factory-simulation tools at the beginning of the product-development process is key to getting the biggest pay-off from this technology. Simulations do not only help ensure that a product will be manufacturable, they also help elicit input from the factory technicians who will actually perform the assembly operations. Showing technicians a 3D, animated sequence from the simulation is an ideal way to communicate about the process, enabling the manufacturing engineers to get feedback from the shop floor before setting up the actual production lines.
To build specialized space-system components, Boeing RSS used Deneb simulation software for production planning and technician training.




By capturing a workable assembly sequence in the initial development cycle, Boeing RSS also greatly simplifies the job of manufacturing planners, the people who plan the work of the factory technicians. "Planners don't start from scratch any more, working from drawings to develop an assembly sequence," Caddick says, "That has already been done in the digital preassembly. They start with the simulation sequence and then optimize the process from there." So what used to take them weeks now takes days, he says.

Finally, creating a pre-defined assembly sequence in software and passing that downstream reduces the learning curve for the people who will actually perform the manufacturing operations. "When we show the simulation to technicians, the initial intent is to get their expert input on assembly processes," Caddick explains, "but we're also showing them something they will be building in the near future. By exposing them to a process that they will have to execute, we reduce the learning curve."

From these examples, it's clear that factory-simulation software has a lot more to offer large manufacturing companies than mere rhetoric about potential returns on investment. For Ford, BMW, and Boeing RSS, the excitement about factory simulation is based on real, rather than virtual, rewards.

Caren D. Potter is a contributing editor of Computer Graphics World.



AutoSimulations
Bountiful, UT
801-298-1398
www.autosim.com

Dassault Systemes of America
Woodland Hills, CA
818-999-2500
www.dsweb.com

Deneb Robotics
Troy, MI
248-267-9696
www.deneb.com

Ames, IA
515-296-9908
www.eai.com

Lanner Group
Houston
713-532-8008
www.lanner.com

Simul8
Herndon, VA
800-547-6024
www.simul8.com

Systems Modeling Corp.
Sewickley, PA
412-741-3727
www.sm.com

Tecnomatix Technologies
Novi, MI
248-471-6140
www.tecnomatix.com