Vehicle Tradespace Analysis
Please enjoy the 2-minute summary gallery below, or scroll further to explore my work in greater detail
In addition, click the Adobe icon below to view or download my design report submission for the 2020 Formula SAE California Competition:
This entry will look different from the rest on this website, because it summarizes a group contribution project that I led. The scope of this project is much larger than the others, and the amount of detail presented is scaled down accordingly. Ultimately, this project encompasses the design ideology behind my last Clemson car, Tiger22, and is the best reflection of the vision I pursued for this team. This page will cover portions of the project that I specifically contributed to alongside portions that I simply oversaw, and I will be clear in distinguishing between the two.
This project resulted from two key objectives of my tenure as Chief Engineer. The first was to shape the team design ideology towards cohesive subsystem integration above individual subsystem optimization. For two years, I laid the foundation of this vision, investing in the simulation and modelling capacity of the team. The result was an expanded library of design tools, which we used to help quantify the effects of subsystem interactions. The final step needed was to combine these tools to re-orient design focus back to the full vehicle scope.
My second objective was to strengthen our influence and value to the community through ambitious collaborative projects with local companies and university groups. I was able to combine these goals when I met Dr. Chris Paredis, the BMW Endowed Chair of Systems Integration at the Clemson University International Center for Automotive Research (CU-ICAR). We shared a similar vision to increase the ties between Clemson's satellite graduate engineering campus and undergraduate students on the main campus.
To meet these goals, we launched a project to develop a full-vehicle tradespace analysis, using the Clemson FSAE vehicle as a case study. This was the final piece I needed to enact my team design vision. I was given $48,000 to assemble a research team for Summer 2019 at CU-ICAR, with full agency to drive the scope and vision of the project. While there, we worked closely with Dr. Paredis, as well as many of his peers and students, who provided advice and guidance throughout the project. The end product was used as the basis of development for the 2020 Clemson FSAE competition car, which was designed from a clean-slate.
The first goal of the project was to define the project objective. Many subsystem design priorities can act in contrast to each other. For example, the powertrain subsystem gains reliability from increased cooling airflow, which in turn can come at the expense of aerodynamic package performance. This defines vehicle design as a multi-objective optimization function, where it can be difficult to weigh the importance of each objective. To reduce the complexity of the problem, we simplified to a single objective: maximize the points scored in dynamic events at the Formula SAE Michigan competition. However, pure performance capacity is not enough to predict competition results, as drive-ability and reliability requirements can be equally critical. To address these requirements while maintaining the single objective strategy, we decided to define constraints that satisfy the criteria while limiting the optimization process to feasible solutions. Where applicable, supplementary simulations could be carried out alongside the optimization to further interrogate top design solutions.
The next step was to determine the scope and complexity to explore. This began with outlining critical systems affecting vehicle performance, and the parameters necessary to describe their behavior. From there, we discussed the most sensitive interactions between these systems, and the parameters necessary to quantify their effects. Starting at the most stratified, we worked to consolidate subsystems whenever possible, in order to manage the scope of the project. The final breakdown is summarized below, with four final subsystems that are joined together via a global "vehicle architecture":
Breakdown of primary vehicle subsystems and critical parameters
Our team decided to build the central tradespace project in ModeFRONTIER, a program specifically designed to carry out multi-objective optimization problems. ModeFRONTIER enables users to link several programs together, such as MATLAB and Solidworks, in a centralized environment, automating the transfer of data between them. From there users can lay out the data flow, and implement various available design exploration and optimization schemes. The team employs a wide range of models, from general MATLAB codes and Excel spreadsheets to specialized CFD and powertrain software. ModeFRONTIER allowed us to link these design tools together, so that changes in one subsystem can automatically propagate to the rest. In addition, ModeFrontier contains several visualization and post-processing tools, which we used extensively to analyze optimization results.
Example of a ModeFRONTIER workflow for a carbon monocoque design exploration, followed by a symbolic representation.
Initially, we struggled to set the scope of the project within a single optimization problem. This was a balance between detail and computational time, and we wanted to strike the right compromise between full vehicle trade offs and allowing for meaningful subsystem development as well. We ultimately decided to break the optimization problem down into four separate levels, named "Optimization Loops", each progressing up the ladder of abstraction. The first loop was the simplest and most abstract, exploring the effects of vehicle dimensions and weight characteristics on global vehicle performance:
As the loops progress, they begin to incorporate greater detail, trading insight for accuracy:
Each loop builds upon the results of the previous one. This allows us to explore every level of detail desired while managing the size and complexity of the optimization problem. In addition, we have the capability to explore certain design concepts without having to start over, or to revisit previous loops to re-examine existing results.
This project required the development of new, dedicated tools to facilitate the integration of the various subsystem models. The first of these was the parametric chassis model, which was developed by our chassis lead William McCormack. This model was an abstract full-vehicle mock up in Solidworks, governed entirely by reference equations. This enabled us to explore various vehicle layout and packaging configurations, including the position and orientation of the driver and other subsystems. More importantly, it constrained our design space to feasible vehicle layouts, allowing us to realistically explore trade-offs between principal inertias and center of gravity location. When implemented in ModeFRONTIER, it automatically outputs the predicted mass distribution properties of the entire vehicle, which feed into the corresponding suspension simulations to examine the effects on performance and maneuverability.
Stylized rendering of the parametric chassis model, displaying critical subsystem bounding boxes with mass representations
Visualization of how various driver position configurations can be utilized (each with their own trade-offs) to achieve a target weight distribution.
The next critical piece was the lap time simulation, which would provide the final overall metric to evaluate and compare vehicle concepts. Since this tool would be used in every loop, it was critical to incorporate modularity and adjustable complexity into the simulation. This was one of my personal projects, and you can read about it here.
The final necessary development was an ANSYS Fluent script (developed by Aero lead Michael Masdea) to automate the meshing, solving, and post-processing of aerodynamic CFD simulations. Aero design explorations were limited to a fixed set of airfoil patterns, using parametric optimization of chord length, placement and angle of attack to achieve target results. The purpose of the script was to carry out a characterization of a given aero package over a range of speeds, ride heights and pitch angle. Eliminating the need for a manual setup was critical step to incorporate aero-related optimization into the tradespace. Initial iterations of the script ran well on personal computers, but several bugs arose when the script was carried out on the Clemson Palmetto Cluster supercomputer. Regrettably, these bugs have not been fixed as of this writing, and this placed a severe road block in carrying out the tradespace to its fullest extent. While progress has been made, it has been very slow, and an exploratory study has begun to consider switching to STAR-CCM+.
Example of a pressure distribution plot automatically generated by the full-car CFD script
Ultimately, the project was limited by the amount of time available over the summer, and when the Fall semester began the time came to focus on finalizing designs for the 2020 vehicle. This meant that the full tradespace analysis was not carried out, but meaningful results were achieved in several areas:
Of the four loops initially planned out, the first two were successfully carried out to completion. What this amounted to was the redesign of our fundamental vehicle architecture, narrowing down to a design concept best suited for our tire and engine selection. This was another personal project of mine, and you can read more about it here.
Suspension and Aerodynamic Platform Control
Bugs in the full-car CFD script kept us from creating proper characterizations of the vehicle's aerodynamic behavior. However, with the information that we had available, we were still able to explore the effects of suspension pitch control on the stability and straight-line performance of the vehicle (you can read about that here). In addition, the framework has been laid for a future optimization once our aero modelling capacity matures. I was actually able to utilize this framework for a separate summer project, which you can read about here to get an idea of what the analysis would have looked like.
Powertrain and Aerodynamics Cooling
This was another project that could not be carried out full due to CFD limitations. However, we still managed to make meaningful progress in this area. Cooling was identified as the most significant interaction between the aerodynamics and powertrain divisions, as it is critical to ensuring the reliability of the vehicle in the endurance event. However, cooling does not directly factor into outright performance, so development went into a constraint function to limit the design space to feasible solutions.
On the team's engine dyno, an engine heat map was created to correlate heat generation to steady state operating conditions. In addition, we combined our cooling fan curve with wind tunnel testing data of our radiator to relate local air speed to the air mass flow rate through the radiator (with fan and shroud). With this information, our powertrain lead, Justin Roberts, was able to simulate engine temperatures through a run on the endurance course using the NTU effectiveness method. This simulation could be repeated several times, varying radiator area or local inlet mass flow rate (assuming even flow across the entire radiator face).
Once a target peak temperature was selected, the simulation could be used to construct the following radiator sizing vs airflow curve:
Around 1 kg/s of air mass flow rate through the radiator (taken at average track velocity), the possible weight savings from downsizing the radiator reach diminishing returns. Based on this relationship, a baseline target of 1 kg/s was selected. However, when mapping several aero design configurations against each other, it became clear that such flow rate was not possible without serious sacrifice to overall downforce, as well as the target aero balance. Lap time simulations confirmed that the performance gains from downforce and balance improvements far outweighed the weight and drag effects of increasing the radiator size, and the flow rate target was adjusted to 0.7 kg/s before selection of the final aerodynamics package.
This plot demonstrates the downforce penalty associated with increasing airflow to the radiator.
Similarly, this plot demonstrated that increasing flow rate while maintaining high overall downforce created a deviation from the target front aero balance of 50%.
Frame Packaging Integration
When frame design is driven by other subsystem development, conflicting packaging requirements can lead to inefficient tube placement and added mass. To avoid this, frame mounting and node locations played a key part in the tradespace analysis.
Significant attention was placed on the effect of suspension pickup point location on both kinematic performance as well as frame weight and stiffness. Rather than letting suspension geometry drive frame design, I generated several suspension configurations that could achieve the same target kinematic characteristics. This allowed us to select a final geometry that maintained high performance without compromising the frame design. You can read more about that project here.
In addition, several structural configurations were generated to explore the trade-off between subsystem weights. For example, multiple concepts were generated and optimized for the rear box section of the frame. The final design shrinks the rear box, which improved differential carrier adjustment kinematics and accessibility while minimizing loads to the chassis. The smaller control arm base meant larger and stiffer suspension links to maintain target deflection values, but this penalty was outweighed by the weight savings in the frame. In another example, adjusting the mounting angle of the steering column increase frame weight, but reduced steering system weight by an even greater amount. Overall, we achieved an estimated 2.5 kg weight reduction simply from shifting focus from individual component weight reduction to total system weight and component interactions.
Pictured here are some of the rear box concepts considered.
Subsystem Performance Optimization
The ModeFRONTIER framework, specifically bolstered by the lap sim and parametric chassis model, created many opportunities to optimize specific subsystem characteristics. The results and design targets from earlier optimization loops were used to guide these optimizations, keeping the design process in the spirit of the original project goal. Had there been more time in the Summer, these would have been incorporated into the tradespace via the higher level optimization loops. That goal has been set aside for future development.
On the suspension kinematics page, you can read more about how I used ModeFRONTIER to inform my control arm geometry design. In addition, lap sim iterations were used to generate suspension loads, which were fed as an input to an automated frame FEA script. This enabled an optimization of tube sizing to minimize peak loads, minimize weight, and achieve target stiffness.
While the full car CFD script never got fully off the ground, the aero division used a simplified version to carry out a parametric optimization of airfoil location and orientation. Those results were used to generate a catalog of front and rear wing configurations to be evaluated in full car CFD. Finally, powertrain was able to use ModeFRONTIER to optimize their intake and exhaust runner geometry to maximize peak horsepower across a target RPM range.
Final Verdict and Future Development
Due to the ambitious scope of the project, the limited time frame allowed during the Summer, and issues implementing the full car CFD script, our team was unable to carry out the project to it's fully intended completion. However, I still consider the project to be a considerable success for the team for the following reasons:
1) The first two levels of the tradespace were successfully carried out, and the resulting vehicle architecture serves as a strong foundation for future vehicle development.
2) Even if the full analysis was not completed, the framework has been set up, paving the way to continue to develop the tradespace. Even though I have graduated from the team, development is still continuing on this project, which speaks to its long term potential and viability. Currently, the chassis and suspension divisions are combining their kinematic and structural models to move into stage three of the optimization loops.
3) Since my graduation, the team has worked on further expanding the scope of the tradespace, incorporating sensitivity to manufacturing defects as an evaluation tool. This demonstrates the potential of the project to evolve and grow to meet the teams needs far beyond it's initial scope.
4) Most importantly, it successfully cemented my team vision to shift focus towards subsystem integration and global vehicle performance. Even in areas where a full tradespace exploration was not possible, the design approach and mindset now exhibits the same goals and values.
Ultimately, the results speak for themselves. At the 2020 FSAE virtual design competition, we were awarded for having the best design in our judging bay of 15 teams. The last time the team was recognized in the top 10% of the field in design was 2012, and I hope this years result serves as an indicator of future success to come.