Engineers at the Fiat Research Center, Turin, Italy, have used innovative approaches and technologies to significantly reduce the weight of an automotive suspension arm. The scope of the exercise was to automatically minimize the weight of the control arm by changes to its geometry - within the design constraints of a given fatigue lifetime and impact strength.
Classically, suspension systems are designed by an iterative process in which computer simulation is used to evaluate each design alternative. Such evaluations often lag behind the current design models, they are often based on rather arbitrarily defined loading conditions, and use only an analysis of the resulting static stress patterns and not the more important dynamic fatigue lifetime evaluation. More importantly, they are open-ended solutions that state how a given design performs - not what the design should be to meet a target performance. Recently, Fiat Research Center engineers used multibody simulation software, with actual road load information as input, to develop subsystem loads that reflect true operating conditions. Then, automated design optimization software was used in conjunction with finite element and fatigue analysis codes to explore the full range of design alternatives for the suspension component. In few days, the software had converged to a design that not only met all durability and impact requirements, it was also considerably lower weight and cost than the original.
Traditionally, suspension components are designed to standards approaching cookbooks that each automobile manufacturer has developed over a number of years. These standards are loosely correlated to durability goals in the actual operating environment: they usually result in a component having to pass a fatigue test that is often too severe. 1This ‘test-based’ approach to durability engineering can therefore lead to over-engineering and excess weight.
Another problem with the traditional design method is the ‘trial and error’ approach to determining the material and geometry to meet the durability specification. Computer-aided engineering methods, such as Finite Element Analysis, have made it possible for engineers to accurately predict stresses and strains under defined conditions, but many do not realize that it is still difficult to use these results to determine where the product will break - and when. In other words, whether or not a component will pass the durability specification. In addition, the current design process consists of developing a concept design, modeling the design, analyzing the model and starting all over again in an effort to solve the problems revealed by the analysis. The results of the analysis tell the engineer that there’s a problem - but provide little or no direction on what is needed to resolve it. While the engineer can determine over time whether or not the design is improved, the potential for actually optimizing the design is little greater than using the traditional build and test method.
Overdesign
The result is that, even with computer-aided methods, there is also a tendency for overdesign, and the suspension is still heavier and more expensive to build than is necessary. In addition, design lead-times are relatively long because several prototype iterations must still be built and tested before engineers can be certain that components will meet durability requirements. In an effort to overcome these problems, Fiat Research Center recently led a three-year program, partially funded by the European Commission, designed to develop new suspension design methods. The method would use a ‘virtual test track’ to accurately predict component loads, finite element analysis to calculate the resulting stresses and strains, fatigue prediction software to calculate the fatigue-life, and an optimization module to close the loop and search for the best design. Of course, the final design would still need to be validated by test, but the intermediate stages would be minimized and automated wherever possible.
Project Leader Dr. Kamel Bel Knani, of Fiat Research Center, began by collecting road load data by physical and virtual experiments to represent a reference mission for the vehicle. To reduce unnecessary computation time, sections from the load time histories that were not significant (from a damage standpoint) were removed. LMS TecWare was used to automatically eliminate the small load cycles from the load time histories without changing the damage content or the phasing between channels. The latter is important to maintain because of the frequency response of the mass-spring-damper nature of a suspension system. The active loads and kinematic constraints were conceived so that load frequency could be increased by replacing the rubber bushing in the subsystem by a metal fixture. This considerably reduced the cycle time required both for fatigue analysis and to perform validation bench testing.
“Real-life” Loading Calculation
While a multi-body program was used to exercise the vehicle on a ‘virtual test track’ in order to determine the loads at the component level within the suspension unit, MSC/NASTRAN finite element analysis code computed the elastic stress distribution for a set of unit loads. LMS FALANCS then used the ‘real-life’ loading calculation and the unit-loads stress estimation to predict the fatigue lifetime of the component. LMS OPTIMUS integrated these two analysis tools, closed the design loop, and drove the product design towards an optimum solution.
The first step was to manually step through each stage of the process to verify its effectiveness. Finite element analysis was used to compute the elastic stress distribution for each load exerted independently. Then load-time signals were assigned to the corresponding finite element model load cases within the fatigue analysis software. The next step was assigning the solution parameters and materials to each element set and performing the fatigue damage calculations. FALANCS uses a multiaxial approach to calculate the local stress-strain path. Material parameters necessary for fatigue damage calculation were estimated from the usual uniaxial tests on specimens. Finally, the computed damage distribution of the component could be visualized using the post-processor of the finite element analysis software. In a separate exercise, HKS/ABAQUS simulated the component behavior under impact loading, resulting in the calculation of maximum reaction force and an appropriate kinematic quantity governing failure under large deformations.
The team correlated these results with bench testing, validating the first part of the computer-aided optimization approach by demonstrating the ability to assess the fatigue performance of a prospective design early in the design cycle. But they were still left with the more interesting task: to determine the optimal component design with regard to durability and impact strength without the usual trial and error design iterations. It was decided to automate this process by using OPTIMUS, which not only integrates nearly any type of simulation software, but can also automatically drive them to explore the design space and search for the optimum design within given constraints. This automation and monitoring of the whole process eliminates one of the most tedious tasks in design engineering.
Bel Knani’s collaborator, Dr. P. Bologna, set up a sequence of analysis functions for a suspension arm that followed the simulation methodology described above. He used a graphical network display to interactively define the inputs, outputs and intermediate procedures during the optimization process. He then viewed the input files of the analysis programs in the graphical display of the optimization software and marked the design variables that he wanted to control. He selected five design variables that the program could change in order to optimize the design including three flange widths, a web thickness and a hole diameter. He also selected four decision responses: the weight of the arm, the fatigue life, the highest reaction force, and an appropriate strain quantity governing impact failure, using similar simple operations guided by intelligent parsers. The arm weight is used as the objective function to be minimized, while suitable constraints were set on the other three decision responses to represent design requirements on arm performance under fatigue and impact loads.
The design of experiments method within OPTIMUS was used to explore the design space and to select a number of virtual experiments that captured interdependencies and reduced the number of required iterations. The optimization software then performed virtual experiments to assess component behavior under fatigue and impact loads. OPTIMUS constructed a response surface model (RSM) to approximate the behavior of the suspension arm when the design parameters vary within their constrained range. An RSM consists, for example, of a quadratic polynomial that is used to approximate the shape of the exact response function. This made it possible to quickly estimate the system response without having to re-run the full analysis sequence.
When the optimal parameters were identified by the Response Surface Model, the design could be analyzed by a formal CAE run. In a sense this helped ‘calibrate’ the RSM by a virtual test: the datum was included in the RSM data set and the RSM regenerated. The cycle was repeated with new
design space bounds, and the model updated with another optimum design for the current model state. The program proceeded through several more stages of analysis with progressively smaller move limits until it reached the true optimum, a combination of design parameters that ensure minimum weight while satisfying the performance constraints.
This new method has the potential to dramatically improve the suspension design process. In today’s highly competitive marketplace, market pressures demand faster product development cycles and continuous performance improvements. The techniques developed by Fiat Research Center in this application can create a virtual design environment in which users can rapidly iterate to an optimized design. The results achieved by Bel Knani’s team are typical of what can be achieved with this new approach - producing better designs in a fraction of the time required by conventional methods.
