How to Roll It Right the First Time

By DE Editors

In the mid-1990s, Corus Construction and Industrial (CCandI), launched along-term project to develop a simulation system that would help it improve itshot rolling process for manufacturing a wide range of steel sections. The goalwas to achieve “right first time” design of these sections for customers in thesteel stock, shipbuilding and construction industries; to supply new steel fasterand more cost effectively.

CCandI, a division of Corus (formed by the 1999 merger of British Steel andKoninklijke Hoogovens, of the Netherlands), a global supplier of metal applicationsfor the construction, automotive, aerospace, rail, and energy industries, wasdetermined that its experienced mill designers would use state-of-the-art computersimulation to run section designs through virtual hot rolling trials. The actualprocess feeds stock like cast slab, bloom, or billet material through a numberof mill stands for shaping into long products such as beams, channels, rails,and pilings. But because the company’s mill designers and engineers had no FEA(finite element analysis) or simulation experience, and because no one had apparentlyever built a simulation system specifically for roll-pass designers, choosingthe right program was critical.

CCandI Product and Process Development Manager David Jennings was confidentit could be accomplished. He considered a simulation system for roll-pass designto be the linchpin of efficient product development. It would help him consolidatedesign resources, institutionalize design knowledge, and make analysis timelyand useful for designers located in the Scunthorpe, England, headquarters.

Since all finite element modeling and analysis of CCandI roll-pass designshad been performed by the company’s Research, Development and Technology (RDandT)department, Jennings turned to the department’s analysis expert, Didier Farrugia.Between 1994 and 1996, Jennings and Farrugia carried out a series of feasibilitystudies to develop the basis of a robust roll-pass design system. When their studieswere complete, Jennings sought support for a roll-pass design project, claimingit would pay for itself in four years. They got approval and moved ahead.

In their studies, Jennings and Farrugia had investigated FEA codes that wouldbe powerful enough to model any rolling process yet flexible enough to allow acustomized interface for FEA neophytes. Farrugia, who relied regularly on bothABAQUS/Standard and ABAQUS/Explicit analysis packages from ABAQUS, Inc. (Pawtucket,RI; abaqus.com) along with codes developed in-house to solve nonlinear problemsin metalforming simulation, learned that the local ABAQUS office was looking fora sophisticated customization job to test out the ABAQUS/CAE pre- and postprocessinginterface. Shortly, a partnership was struck.

“We considered many different options for this system,” says Farrugia. “We decided,since ABAQUS was our primary solver code, that a tight coupling with the pre-and postprocessor was the right way forward. There was an element of risk forus because so much of the technology had to be developed. But it was good timing,and I knew we could trust ABAQUS for strong professional engineering and technicalbacking. There are not a lot of organizations like them in the technical world.”

ABAQUS committed its FEA consulting power and technical know-how to the project.Working closely with Farrugia, ABAQUS developers in the U.S. extended the software’scapacity to simulate the key physical phenomena of hot rolling: large deformationof the rolled material; strong coupling between thermal and mechanical response,contact, and friction; and cooling of the material between roll passes. At thesame time, ABAQUS UK focused on tailoring the user interface for the productionenvironment of mill designers.

“High-temperature steel is very complex,” says Farrugia, “with competing behaviorsof hardening and softening mechanisms. Corus has a massive background of understandingof material properties, and we knew the best way to tackle 3-D hot rolling simulationtemperature coupling. ABAQUS helped us harness the engine of our analysis expertiseso designers could drive it.”

Several advanced features for modeling hot rolling were added to the ABAQUS solvercodes. Among them was an innovative mathematical approach to update the mesh duringsimulation. Two phenomena that occur during physical rolling of metal are elongationand spread. The difficulty for analysis is that the meshed workpiece is subjectedto elongation and spread during simulated rolling, so the mesh can become heavilydistorted and lose accuracy. At the time, ABAQUS was working on a method calledadaptive meshing, which smoothes mesh distortion during each roll pass. With Farrugia,the developers decided that this was the most efficient technology for handlingsingle-pass 3D rolling simulations.

“Localized deformation is so intense that the mesh degrades,” Farrugia explains.“Using a Lagrangian formulation the material and the mesh are bound together,so element distortion will affect the material response to the extent where theanalysis may produce inaccurate results or fail. The Arbitrary Lagrangian-Eulerianmethod now ensures effective mesh smoothing within the rolling pass, and staticremeshing is carried out at the interpass time between mill stands. It was thekey to accurate spread predictions.”

Another feature was a new algorithm for automatic mass scaling in ABAQUS/Explicit.Farrugia wanted to scale the mass of the model automatically to reduce computationaltime. The danger is that too much mass scaling will result in the wrong inertialforces being applied during rolling. A new physical algorithm for calculatingand applying the time incrementation during rolling simulation in ABAQUS/Explicitnow ensures accurate solutions for any scaled section models, with up to 100 timesfaster processing in the CPU.

Further reduction in runtime was achieved by a third advancement called steady-statedetection. Rolling is a steady-state process. During simulation, product exitseach rolling operation at a location called the exit plane, where it reaches anear-constant shape before continuing to the next set of rollers. At the exitplane, the analysis system monitors model parameters such as roll reaction force,spread, torque, and plastic strain. When these parameters satisfy preset criteria,the analysis stops. The system then automatically extracts a 2D steady-state cross-section(and its internal state, i.e., temperature), runs an interpass cooling, and remeshesthe model for the next rolling pass. The system permits steady-state detectionfor long products produced by as many as 45 roll passes, with a CPU time rangingbetween 10 minutes to a couple of hours on an SGI Origin 2000 system.

Hot rolling of steel sections is an engineering specialty and, traditionally,roll-pass design engineers acquire their expertise over time. They are responsiblefor determining the most efficient mill configurations so the end product hasthe correct geometry. It involves predicting how the metal will respond as itpasses through rolls and selected grooves at different forces, temperatures, andspeeds, and ties together the electrical and mechanical constraints imposed bytheir mills.

The challenge to Jennings and Farrugia was combining that individual knowledgewith the analysis expertise of RDandT to arrive at a robust, user-friendly,and tailored simulation system. They had to figure out how to translate the milldesigners’ hard-won empirical and implicit knowledge into specific, recordabledesign rules that could then be formulated into software. But far removed fromthe creation of the FEA routines was the challenge of convincing the designersthat their process, which was so dependent on hundreds of individual skilled judgments,could be successfully automated.

“The design engineers were skeptical at first,” Jennings admits, “as they didnot believe that complex models for rolling would be useful for design, despitethe successful outcome of the feasibility studies. They were put off by the languageof analysis.” Nevertheless, he began to ask them questions: If you were able todo your own analysis modeling and process simulation during roll-pass design,how would the system need to work? What kind of information would be useful tohave? How should it be presented?

Over the course of several weekend retreats, the design engineers came togetherto address these questions. Later, the modeling experts from RDandT joined in.“In general,” says Jennings, “the roll-pass design engineer has to contend withthree governing factors: the type of feedstock available, the intended finishedproduct, and the number and design of the mill stands available. The goal is tomanage these constraints to get the product you want with optimal rolling yield.There are some common principles in rolling, but also many different product-and mill-based constraints, and rules for specific designs, such as a rail ora piling. These specific principles, and the reasons behind the designers’ rules,were what it was necessary to capture.”

It was vital to get the principles right. “Rolling yield is important,” Jenningscontinues. “A one-percent improvement on only one of our mills translates intoa savings of considerably more than £1 million ($1.75 million) per year, whichis a prize worth having.” And rolling yield is directly related to the robustnessof design. “The rules for creating robust roll-pass designs carry allowances forthe physics of real life,” says Jennings. “Let’s say we determine that a bar shouldcome out of the furnace at 1250 degrees C and should be 800 degrees by the timeit’s finished. Ninety percent of the bars will follow this pattern. But what ifsomething happens in the mill, and the bar is delayed so it’s standing in themill for two minutes? The temperature will be lower, but if the design is robust,the bar will still roll successfully.”

Jennings was certain the roll-pass design system would increase design robustness.Before long, the designers also recognized this. As work on the system progressed,the designers tested every new simulation tool developed by the RDandT teamworking with ABAQUS.

Jennings and Farrugia coined two watchwords for development of the user interface:no shortcuts. Everything depended on making simulation friendly, robust, and transparentto the roll-pass design engineers. With the help of a third-party consultant,they developed and tested a detailed design specification for the interface, gatheringfeedback from the designers through paper-based mockups, manual usability trials,and videotaped role-play exercises. Then, using a scripting language called Pythonthat is embedded in ABAQUS/CAE, they created a user-friendly front end for settingup analysis models and for obtaining results.

As the specification required, analysis happens right at the CAD workstationsthe designers use for roll and guide drawings. Design engineers themselves setup a single or multipass schedule for a rolling simulation. From a limited seriesof selection screens, they can specify the sequence of mill stands, rolls, orgrooves; the mill layout; the temperature at the start of rolling; gap settingsbetween rolls; feedstock; number of passes; type of analyses (isothermal, thermomechanicallycoupled); and so on. The simulation itself is completely automatic and invisibleto the design team, performed by ABAQUS implicit and explicit FEA code directedby the customized scripting.

“The system computes the shape evolution from ingoing feedstock to finished product,“says Jennings, “which is exactly what our roll-pass designers need. It allowsthem to explore the rolling process incrementally, pass by pass, so they can makesure material deformation proceeds as planned for each specific product. And itgives them the opportunity to do what-if studies to see if the design will holdup to real-life conditions. What if the bar is a little bit colder? What if there’sa delay after the third pass? What if the rolls are offset? Because they can testthe rolling process virtually, we don’t have to do as many mill trials.”

According to Jennings, those mill trials have decreased dramatically since theroll-pass design system was launched in 1999. “We now introduce 75 percent ofour products into the mill without a physical trial at all,” he says. “We alsohave major success in improved mill yield and throughput per hour. And on oneof our rolling mills this year we’ve eliminated 90 roll changes, simply by rationalizingdesign and shortening the mill. The system greatly improves mill availabilityand productivity.”

Since 2000, Corus has added a guide and roll-stress module to the system, alongwith capabilities for design of experiments and, shortly, design optimization.Ongoing developments include tools for roller straightening, material microstructure,friction, tension and looper control, and porting to AutoCAD and other CAD packages.The designers can run simulations on their laptops if they need to work remotely.

And that four-year time frame Jennings estimated for the system to pay off? Theproject actually took only about half as long to reach payback.

“We are stretching our mills to their physical and creative limits,” says Jennings.“Simulation gives the design engineer freedom to try out radical solutions. Youcan cheerfully test some novel idea in absolute safety and security, watch whathappens in the system, and learn something about the process that you can lateruse.”

“It was a gradual building of confidence,” he says, referring to the roll-passdesigners his team worked with. “Now those same people would not do a design withoutusing modeling and analysis. They have absolute confidence that what they seein the simulation is what they will see on the rolling mill.”

Send comments about this article c/o Desktop Engineering Feedback.

ABAQUS, Inc.