A manufacturer of drivetrain components for heavy equipment uses design optimization software to develop products more quickly without the need for repetitive and costly prototype testing.
Traditional approaches to product development require considerable time and effort to design, build, and test prototypes. But design optimization software programs, coupled with finite element analysis (FEA), are helping engineers to shorten the process.
For example, Engineers at Rockford Powertrain Inc., a leading supplier of drivetrain components to heavy equipment manufacturers such as Caterpillar and John Deere, use design optimization software to substantially reduce the need for prototypes and laboratory tests. The result: better products brought to market in less time, which translates into faster response to customer needs.
Before incorporating CAD software into its engineering process, Rockford’s senior product engineer, Gordon Cummings, put a design software package from Algor Inc. to the test. He evaluated the software capabilities by using it to analyze an end yoke, a drivetrain component popular in mining, construction, and agricultural equipment, that must withstand a variety of forces at constantly changing angles.
Some of these yokes are made from steel, not for strength, but to facilitate welding to tubing. Other types, called fitting or slip yokes don’t require welding. Thus, they can be manufactured from less expensive ductile iron. The ductile iron yoke selected for the software evaluation is a U-joint component in the driveshaft of a log skidder, Figure 1. In this application, the U-joints are exposed to considerable abuse as the skidder drags logs over rough terrain to a bark stripper. Because the yokes had been tested in the laboratory and experimental data was available, this data provided a good opportunity to validate results of the FEA program.
Results of the software analysis confirmed the earlier laboratory test results. Mr. Cummings commented, “A comparison to the parts from our fatigue testing machine shows failures along the clusters of finite elements with the highest stress levels in the software model.” Figure 2 illustrates the yoke stress pattern produced by the FEA program and Figure 3 shows a yoke that was deliberately broken in the company’s test laboratory at Rockford, Ill.
How the software works
Based on the satisfactory results of the yoke analysis, Rockford expanded its Algor software package to include several complimentary programs and began using them for both design optimization and failure prediction.
The design process starts with a Superdraw program, which enables users to create models of parts or import them from CAD packages. In either case, the size of the part is established by the scale drawing. At Rockford, an engineer typically imports a drawing from AutoCAD into the Superdraw program, then applies a finite element mesh to the part surfaces, using Supersurf. Then a program, called Hypergen, automatically generates a solid mesh of tetrahedronshaped elements (like a pyramid) based on the surface mesh.
Once the model is ready, the engineer adds boundary conditions and loads. This is where experience and good engineering judgment is essential to obtaining an accurate simulation of the part. Finally, a Stress Analysis program, as the name implies, calculates stresses in the part and pinpoints the probable failure location.
Though it takes 2 days or more to create a model, today’s finite element analysis software performs an analysis in 5 hr or less, depending on model complexity. Prior to using Hypergen, converting the model from a surface mesh to a solid mesh for analysis took a week or more.
Three years ago, a finite element analysis typically took about 8 hr to run on a 286 PC. To avoid a long wait, the engineer often left the PC on overnight, anticipating a solution the next morning. But more often than not, someone turned the power off or an electronic gremlin interrupted the run. Today, a 486-66DX PC with 8 MB of RAM and a 750 MB hard drive completes a comparable solution in 10 to 15 min.
Historically, engineers at the drivetrain manufacturer used traditional methods of calculating product parameters, such as stresses and endurance limits, then tested the product to verify the calculations. Their test facility has several test machines that generate torque and speed, one of which can spin a 50-lb part at up to 12,000 rpm to determine its burst speed.
Continue on Page 2
Discrepancies between calculations and test results were common for complex parts, because features like notches, holes, bores, compound curves, multiple loads, and centrifugal effects made accurate calculations and failure predictions difficult at best. When a tested part was not strong enough, the discrepancy was usually fixed by adding material to strengthen the part. This process, of course, seldom results in the removal of material because if it doesn’t break, you don’t fix it.
In the last four years, Rockford used the FEA software to improve the design of several drivetrain components. For example, they analyzed a 65-lb stub yoke to determine if they could reduce its weight without affecting its service life. They were able to shave 19 lb from the yoke, thereby reducing not only its cost, but also the inertia of the drivetrain.
In another case, a customer requested a radical change in a bracket supporting a fan drive so they could relocate the cooling fan. Engineers were concerned that a cast iron plate, on which the drive was mounted, would deflect under the offset load. The project engineer considered adding cast ribs on the plate to increase its stiffness. However, an FEA analysis showed that the plate was adequate without adding stiffeners. This provided substantial savings in both test and manufacturing costs.
In a third example, a power takeoff assembly was vibrating excessively. Through laboratory tests, engineers traced the problem to shaft deflection caused by a clutch (on the shaft) that did not apply uniform clamping pressure when it first engaged. To duplicate this situation, they built an FEA model of the shaft and tested several load variations until the deflection matched that measured in the lab. Using this model, they optimized the shaft design by adding material in the deflection area, thus eliminating the vibration.
Recently, the company used FEA to predict failures in a number of applications. One case involved a manufacturing fixture that failed. Engineers analyzed the part in only an hour and were able to verify not only the location of failure, but also the angle of the fracture. A testing process would have taken several days. Based on this analysis, the part was redesigned and has been operating trouble-free.
In another case, the engineers conducted an FEA analysis to determine the effect of putting a proposed vent hole in a torque-carrying tube. The analysis took only 15 min. Then they used a program called Pizzaz to generate a graphic fourcolor print showing the adverse effect of such a hole. This program, from Application Techniques Inc., Pepperell, Mass., allows you to print the results of an analysis in color.
Other parts analyzed include cast iron brackets, ductile iron yokes, steel shafts, stampings, and sheet metal parts. In almost all cases, the program accurately predicted the weak link. It even predicted within a 5% variance the shearing of teeth from a clutch plate.
Gordon Cummings is sold on the use of FEA, and its ability to save money. Though specific savings are difficult to quantify, he estimates that the complete hardware and software package, which cost about $6,500, paid for itself in the first year.
Engineers still needed
FEA software excels at handling tedious number-crunching operations and providing analytical solutions for complex parts. But, experienced engineers are still the key to successful design optimization.
Matching an FEA model to the real world is a challenge that requires sound engineering judgment. For example, when a part is symmetrical, designers often model only half of the part. But, if not properly constrained, the resultant model may not react to load in the same way as the actual part. In such cases, it may be better to model the entire part. Boundary conditions (anchors) can constrain any combination of six degrees of freedom and must match those constraints on the actual part. “I can’t overstress the importance of properly applying loads and boundary conditions, and interpreting the results,” says Gordon Cummings. “FEA is an easy-tomaster tool. But, it must be mastered.”
Rockford engineers discovered a valuable side benefit to using FEA software. When a part with several highstress points is broken, the fracture propagates from one specific highstress point to another, depending on the direction of applied load. For example, loose bolts in an assembly can alter the normal load direction and thereby cause the crack to propagate in a different direction. By analyzing the effects of such factors (loose bolts) on the computer, an engineer can often determine what caused a part to fail by observing the direction of crack propagation in the part.