Conceptual and Objective Design in FSAE
“Typical engineer. He thinks up a single framework for the overall solution, and won’t look at anything else.”
- UMich industrial psychologist in casual conversation
Why Conceptual Design?
One of the axioms of engineering design methodology is that there are (very) many feasible overall frameworks (or conceptual designs – n.) which are well-suited to solving the same design problem, and that choosing the best of these (Formulaistas tend to call this the “optimum”) is difficult and non-obvious. The process of conceptual design (v.) deals with identifying and selecting the most promising conceptual design (n.), before getting down to defeating the devil in that concept’s details (its ‘embodiment’).
Most Formula teams have about the same design goal – to design and build the best racecar that they can, subject to whatever constraints they are working under. This is often approached as incremental design – fixing what went wrong last year and improving those systems of that old racecar concept that seem ripe for it. This cautious approach usually leads to some degree of overall vehicle improvement. But it misses a high-return design opportunity: by applying the understanding gained from building and racing versions of that old concept, a new overall design concept can be developed that is better targeted for racing success, and reaches that target sooner.
Further, drafts of the new scoresheet coming for Formula SAE emphasize the idea of overall vehicle design, arching over the independent design of vehicle systems. Focusing too much on systems, and ignoring overall design, will risk a design event point deficit. Overall design (and the process for achieving it) emphasizes overall performance of the racecar and the choice, integration, and arrangement of vehicle systems to best enable this. These ideas are addressed by the engineering discipline of conceptual design.
The process of conceptual design is fluid, depending on the domain to which it is applied. The purpose of this article is to explore the application of conceptual design to FSAE racecars. The goal is for teams to develop better racecar concepts, and to do so more quickly and accurately.
How Conceptual Design Works in FSAE
Design looks for an optimum – the machine that can perform its functions best. The perfectly accurate way to find an optimum is to completely design and build all of the reasonable design variants and then run them against each other – i.e. flood the design space (the set of potential feasible designs) and test to validate. This is obviously impractical.
But consider the Pareto principle – the 80/20 thing: by committing only 20% of the design effort that would be necessary to completely design each variant, can we predict its performance with 80% accuracy? If so, then more variants can be considered (although more approximately) and evaluated (with some loss of accuracy) in the design effort, and the likelihood of locating a more optimal design variant is improved. This is a common feature of conceptual design: quickly and approximately sort through partially defined design variations (in FSAE: the overall vehicle concept) to indicate the most promising design, before committing to the huge effort of an embodiment design (in FSAE: preliminary and detailed vehicle systems design).
FSAE fits neatly into the Pareto principle. Lap time simulation (using tools developed from good vehicle dynamics and probable course layouts) certainly exceeds 80% accuracy in performance prediction, and the inputs required by lap time simulation (mass properties and summary characteristics for axial and lateral thrust) are far less than 20% of the total design content of an FSAE racecar.
Comparative lap time simulation (concept evaluation) over a range of input sets (conceptual design variants) is the first great tool for FSAE conceptual design. Lap time simulations for different events can also be combined to create potential total points in FSAE running events. After using this method to canvas the design space and determine an optimum vehicle concept, embodiment design follows to complete the design and verify the concept, iterating back through conceptual design as necessary.
FSAE Conceptual Design Tools
Generative tools are needed to create an efficient range of potential design concept variants. Evaluative tools are needed to examine the relative rank and multi-faceted goodness of these variants, and point out directions for improvement. Before the process of conceptual design can be followed (sweeping the parameters and plotting success), the tools that guide this process must be developed and calibrated.
Lap Time Simulation
In specifying a concept design variant for evaluation (lap time simulation, etc.), it is important to keep the lap time simulation input set as general as possible. Rough thrust curves (thrust at just three or four speeds) are part of this set – they are necessary to make axial acceleration. But specific engine or motor choice is not – this is the means for later fulfilling (embodying) the requirement represented by the rough thrust curves. Tire coefficients and weight transfer characteristics (again, at just a few key speed and radius points) are in the simulation input set – they make lateral and yaw acceleration. Damper choice and suspension arrangements are not – these are the embodied means. In conceptual design, it is important to work with as small a set of inputs as necessary to drive a useful lap time simulation. Generating more input information, since total design time is limited, limits the number of conceptual design variants that can be considered, and makes the optimal concept easier to miss.
For example, for thrust curves for conceptual design variant evaluation, it is sufficient to use representatives of what might reasonably be achieved by broad classes (one cylinder or four; one motor or four), forecasting what the Powertrain Team’s best efforts might be expected to achieve during embodiment design with engine/motor development, driveline specification, squat, and tire coefficients. It is not necessary to use the full torque characteristic of a specific engine or motor. If such a characteristic were used, then the freedom of conceptual design would have been limited.
Not all performance evaluation tools are lap time simulations. Performance evaluated only by lap time simulation often fails to point out what is most important in a concept, as well as how that concept might be improved. Consideration should be given to keeping a variety of performance evaluation tools, from full Autocross simulation down to Acceleration simulation (Matlab or spreadsheet), Skid Pad simulation (hand calculation), or even simple corner entrance braking point (with both axial and lateral components). Plus everywhere in between that is useful.
An accurate evaluation (or lap time simulation) tool can be an invaluable aid to excellent design. An inaccurate tool can be misleading. For instance, a bicycle model will say a few useful things about how fast a racecar might run Autocross. But since racecars (especially FSAE racecars) do not turn as bicycles do, simulations that do not accurately represent four-wheel dynamics might not only result in misevaluation, they will also not fully reveal the performance impact of alternate suspension systems and their associated masses and costs. Combustion engine models without gas dynamics and electrical powertrain models without thermal effects lead to similar difficulties. A good design engineer always approaches simulation tools with some skepticism and honest questioning, striving to understand the insights and limitations intrinsic to any evaluation tool.
Mass Model
Each set of lap time simulation inputs must represent a feasibly buildable racecar. It is silly to draw the conclusion, from simulations, that doubling the thrust and halving the mass improves performance, if such a racecar cannot be built. Each lap time simulation input set must be buildable within the constraints of FSAE Rules, available money, available facilities, acceptable labor, and the physical reality of total vehicle mass. Mass is one of the key performance-determining parameters for FSAE racecars, and the mass in a lap time simulation input set must realistically correlate with the other inputs. For instance, If the thrust curves are raised, then the mass will surely also increase. This brings us to the second great tool for FSAE conceptual design – the mass model.
A mass model is a spreadsheet of everything on the racecar (judiciously combined as subsystems of several parts) and their individual masses. This should not be confused with the Cost Report’s Bill of Materials (BOM), because the BOM is: 1) done after the fact (of design), and so cannot contribute to conceptual design (which works on the front end of the design process); and 2) the BOM is far too much work to complete for more than one design variant (and so multiple concept variants cannot be evaluated to find an optimum). The mass model must be a Pareto (80/20) construct – applying 20% of the design effort to find (with 80% certainty) the probable result. The mass model need not predict the absolute weigh-in mass, but must predict the relative mass impact of embodying to accomplish different levels of (for instance) lateral grip.
For a given thrust (for example), the mass model must correlate to a final drive system mass (or perhaps use a finer divisions of subsystems, like sprockets and axles). This means doing some example system designs for a range of system input parameters in order to understand how the system design changes (design is not just about getting to the design point – it is also about understanding the sensitivity of that point to variations). Axial thrust will also correlate to an engine or motor mass, and these further correlate to a frame mass. Subsystem masses often depend upon the ultimate total vehicle mass, making the computational structure of the mass model iterative. The subsystem masses add up to a whole-vehicle mass that can be input with the axial and lateral thrust parameters to drive a lap time simulation for an achievable concept variant.
A mass model contains, in its lines of code and set constants, much of what a team knows about FSAE design. What is a reasonable wheel hub mass? How might the mass vary if the motors are smaller (or inboard, or with higher reduction ratios)? How much less might it be if the embodiment designers use a good level of structural optimization? A good mass model is tough to build but enables vast predictive ability. It can also be a powerful way to build design teams, through the effort to acquire and integrate design information for correlations. The model is a growing thing, as the team learns more and more from year to year, and stores their knowledge in their mass model. With good documentation, it can be a vessel for design information continuity from year to year. It can generate system and part goals (how much mass are you willing to allow to the upright designer?). A well-maintained and calibrated mass model grants an FSAE team the ability to design for truth, as opposed to designing for hope.
Mass Model Extension to Cost and Production
The mass model may also be extended as cost and labor models (cost for conceptual design purposes, and not cost to the precision of the Cost Report). Cost should consider overall program cost, including shop tools and supplies, testing expenses, design risk (repair, redesign, remanufacture) and competition travel. It is important for cost-constrained teams (i.e. all teams, though some more than others) to strive to find the concept that will achieve the best performance within their limited cost. Finding this finds the optimum conceptual design under the constraint. This is much more effective than simply rejecting specific high-cost system choices (like buying awful tires because they are cheap).
A labor model (unit: labor time) and production schedule (unit: calendar time) can also be worked as an extension of the mass model. Additional items need to be included for: system assembly; system test; system rework; overall assembly; overall test; and the usual test-tune-train cycle. For production, the model should include: total task list; task prerequisites; task dependencies; task costs (labor, money, facilities); task durations; and the tangible result of each task (its hard deliverable item). Labor and production can also account for fabrication training time that might be associated with particular system choices. Modeling production along with conceptual design makes it easier for a team to design up to their production potential, instead of being limited by their production fears.
Conceptual General Arrangement
A full set of mass properties (for lap time simulation) requires center of mass in three axes as well as yaw gyradius. Estimating these parameters requires an approximate general arrangement, as racecars are constrained to be assemblies of non-intersecting volumes. This general arrangement should be done in CAD (for transferability, adjustability, and computability), but it must also be very rough. Its parts should be blocks of rough shape, without detail (though the block placements should approximate the conformal inter-system fits that should be achieved in later embodiment design, instead of giving every block an over-wide safety buffer).
Conceptual design CAD arrangement must be rough in order to be quickly made in order to cover more design space within limited design time. The center of mass of each block may be approximated as its center of volume (or placed at some individually referenced input point, if known), and block gyradius may be approximated as the gyradius of an equivalent brick (unless better information is available). Then the whole-vehicle mass properties may be calculated. Lap time simulation dimensions (wheelbase, track widths) must also be checked and iterated through the mass model.
The process of general arrangement of conceptual design variants is fun. This, after all, is doing the layout design of a racecar. This fun will lead to more fun (i.e. successfully racing the optimal racecar) if the various layouts are skillfully made and enable a truly selective design (well-advised selection from among a promising set of variants). Some arrangement advice:
A well-known truth of FSAE design is that all masses should be small, low, and centralized. Small mass – that’s obvious. As low as possible (including the driver) to bring the VCG down. Packed toward the center of the wheelbase to reduce yaw moment of inertia and enhance agility.
Typically, small is light. Parts and systems should be as small as they can be, and still accomplish their functions. If some parts and systems can accomplish multiple functions, then other parts may be removed – infinitely lightening them. And the only functions that matter are those that get points in FSAE competitions.
It is best to leave the frame or tub until after the other subsystems have all been placed relative to each other. Place the systems where they need to be, dimensionally, for purposes of function and packing. Then stitch the frame around them. If this results in design difficulty for the frame, then resolve this by iteration of the system locations. If the frame is sketched first, then whole-vehicle size and mass are likely to grow – irreversibly so. Many FSAE designs are ruined by placing systems on a frame that carefully allowed more than enough room for them. The systems then expand to fill the frame, and so none have a chance of becoming smaller and lighter.
The driver should be modeled as a part for both Ms. Petite (5%F) and for Mr. Beefy (95%M). This is per FSAE Rules. The driver space (minimum volume for the driver to perform the driver functions) should also be CAD modeled as a massless volume, checking the rough feasibility of each extreme driver’s biomechanical functions (sight lines, pedal force, steering torque, etc.). ‘Small is light’ applies strongly to this volume, since it impacts the volumes of other systems and so multiplies the lightness effect.
Aero can be amazing – the performance difference between aero-on and aero-off for a good aero racecar can be stunning. But too-simple aero performance predictions, not supported by painstaking test, usually result in over-prediction of downforce - sometimes grossly so. Experience and good FSAE engineering judgement, including cost and labor estimates and close correlation with physical test, should accompany any decision to include aero in a conceptual design variant.
Teams should add their own pearls of FSAE design wisdom to these!
Other Design Considerations
There are other non-performance design considerations besides cost and production, such as: drivability; reliability; reparability; parts availability; manufacturing risk; and others. The broader world of engineering addresses quantification of the design impact of each of these. But accurately applying this knowledge to a practical FSAE design effort can be challenging. This is not to say that these concerns should not be included in evaluation of a conceptual design variant. The classic technique is the weighted decision matrix. The value of each consideration (performance, cost, drivability, reparability, etc.), for each variant, is scored (either quantitatively for performance and cost, or qualitatively where models and data are lacking), multiplied by a weighting factor (unique to each consideration), and summed to compute a total satisfaction score for each variant. Comparison across all variants indicates a solution.
Classic decision matricization creates a soft design solution of sorts, though not necessarily a good one. Much of the design methodology community actually sees decision matrices as futile, since it may never be possible to accurately score and weigh each consideration relative to each other. However, if the decision matrix is treated merely as an informational summary (all the attributes of each variant are summarized on a single page), from which a qualitative decision of engineering judgement can be made (instead of just going with the calculated satisfaction), then the decision matrix has value.
Continuous Product and Process Improvement
Vast design information and great competitive advantage are yielded by these models for performance, mass, arrangement, cost, and production. But only if these models are appropriately accurate, and the designers understand both the strengths and the weaknesses of their models. There is the story of one recent Formula team whose Acceleration event competition goal was 5 s. Their simulations suggested that all of their conceptual design variants were capable of 4.2 s. And so no further effort went into powertrain, mass reduction, or launching traction. At competition, they did not make the 4.2, or even the 5. The cause was falling for an overly simplistic (and hence inaccurate) acceleration simulation. The peak motorcycle-rated (i.e. unrestricted) engine power was applied as a constant over the entire acceleration run – not an outcome that vehicle and powertrain dynamics allow. The moral of the story is, at least, to know when a simulation might be inaccurate and to allow for this in design evaluation. And at best, to spare no effort in understanding vehicle dynamics and powertrain simulation, and correlate these with physical test data. Another moral is that really useful simulation is much more for the practical improvement of a team’s design, and much less for merely impressing design judges.
FSAE Conceptual Design Process
Summarizing the elements necessary to specify each potential FSAE conceptual design variant, we have:
Performance simulation input set
Broad choices for configuration or type of each major system
Rough general arrangement to check feasibility and determine mass properties
Summarizing the FSAE conceptual design evaluation tools, we have:
Mass (properties) model correlated to the lap time simulation input parameters and broad system choices
Lap time simulation and other performance models, correlated to real event performance, possibly extended to competition points prediction for multiple running events
Cost model, especially if this is a constraint
Labor and production model, especially if this is a constraint
System for collation and presentation of evaluation process data
Once these necessary elements are in order, the process of generating and evaluating conceptual design variants for the purpose of making an optimal choice can begin.
To select potential conceptual design variants, start with just a few candidates, but cover a wide range. Perhaps consider: a powerful racecar (with higher mass); a light racecar (with limited thrust); a grippy racecar (probably heavy); a cheap racecar (probably light). Perhaps consider a few of your favorites (your last best racecar, Stuttgart’s racecar). Set up the realistic inputs, check the arrangement (iterating between these two), and run the design evaluation tools. Consider the results. These results start to show the shape of design space (multi-dimensional plot of design goodness v. the variables of design). If a conceptual design variant exceeds a constraint (perhaps cost is too high), then adjust the variant to bring it within the constraint. Record the results in a decision matrix, if desired. Then find out whether a new variant selected to be between two points of your initial conceptual design range is better than either endpoint, or worse, or whether goodness varies monotonically between endpoints. The more variant points that are defined and evaluated, the better is the definition of the design surface for FSAE racecars, seeking victory, while satisfying constraints.
This optimization method is sometimes denigrated as a random search. There are many more mathematically sophisticated optimization methods, but they usually work poorly on something as multi-faceted as an FSAE car. The random search method works well if there are few enough variants (say less than 40), that each can be well understood by its characteristics and by its evaluation results. In other words, use judicious reasoning instead of purely random search to guide your choice of potential competition points-getters. And remember, the only purpose for an FSAE racecar is to get points at competition.
Objective Design for FSAE
Still, the above process for selecting conceptual design variants is a little hunt-and-peck: imagining a design variant and then evaluating it. A better way is to be more objective. Objective design seeks to match selected conceptual parameters (such as lap time simulation model inputs) analytically to performance outcomes (such as event times). I.e. analytically associate thrust curves and mass properties with competition point totals (running events).
But this evaluation is fuzzy. Competition results might depend upon: track conditions (temperature, moisture, sandiness); driver skill (the team’s and competing teams); the actual course layout; typical stochastics; and the fickleness of the racing gods. Further, some events are just plain hard to simulate accurately.
Plugging notional lap time simulation input sets into a competition point simulator and tabulating the output might indicate whether one design variant is better than another variant, but it will not predict an absolute event finishing time, and it might not reveal why a good (or poor) result was obtained. There may be no indication of the need for both zero-speed thrust and thrust at some higher speed (like mid-Acceleration event at or at corner exit). From point simulations, it might be hard to pick up the value of quick turn-in or the impact of good brakes.
A better way to find promising conceptual design variants is to perform objective design – find numerical design targets which express the direct physical outputs of the racecar as a mechanical device: its accelerations. Then choose design concept variants such that all considered are capable of achieving these acceleration objectives (or some of them, if performing trade-off studies on the objectives).
The purpose for an FSAE racecar is to win the competition, or to place as highly as the team’s performance-constraint surface will allow. But the mechanism of the racecar - it’s roll axis, its torque bias, its brake pressure – is designed to accomplish something more quantitative. If an FSAE racecar is really an acceleration machine (axial, lateral, yaw, braking), then what acceleration values should the racecar be designed for? At what speeds and turning radii should these accelerations be achieved?
Study of lap time simulation results, past competition results, and the designs that achieved them should yield an understanding of what designed-in acceleration levels are necessary to produce an estimated event placing. Then, conceptual design variants may be pre-tuned to meet these objective acceleration levels. Then, every conceptual design variant that is subjected to overall performance, cost, and production evaluation will be pre-qualified by meeting objective goals for acceleration. A reduced number of better-qualified variants will soak up less conceptual design time. Plus, these objectives are well-suited to become design goals for system designers.
Care must be taken in correlating acceleration objectives with event performance. For example, winning Skid Pad times are always well above what might be expected for the winning racecar’s peak lateral acceleration capability. It is important to understand and account for this deficit. Autocross simulations often cannot account for driveability and driver skill, with the result that higher accelerations must be built into the racecar to make up. Endurance brings the characteristics of mechanical and electrical degradation, fuzzying up the objectives as a result. Objective goals must be thoroughly calibrated with past racecar characteristics and actual performance – the team’s as well as those of any other FSAE racecar for which sufficient data is found.
Further, the value of the whole FSAE design experience (design for excellence) can be disappointed if the objective acceleration goals are set too low. As noted above (on Skid Pad times), setting them to actual event performance will surely yield an under-performing car. A factor of safety (or correlation) is needed. And then there is the matter of how much excellence a team should design its racecar for. Perhaps it is calculated that it took 1.5 g’s of lateral acceleration to win last year’s skidpad. Perhaps that old winning car is analyzed and found to have a peak lateral acceleration potential of 2.2 g’s. Perhaps the current team’s overall goal is to place in the top 5, and for this placement, it is determined that only 1.4 g’s are necessary. The team could arguably achieve this goal by designing for 2.1 g’s. But why not design for 2.2 g’s? If cost or other constraints are a concern, then the cost model should address these and find the lateral acceleration available for cost. If a team fears to venture, then it should ask what, specifically, it fears. If a team doubts its ability, then it is more likely to improve its ability by pushing the performance envelop. Until the team bumps into a real constraint, it is no harder designing for 2.2 g’s than for 2.1 g’s.
Summary
It is often said that FSAE is scary close to the real world of advanced technology product development. It is even said, by some in that world, that FSAE exceeds those challenges. Conceptual design is one of the tools necessary for navigating in that higher plane.
The process of conceptual design for FSAE racecars (once assessment and generation tools have been built and calibrated) begins with generation of a range of conceptual design variants. Each variant should be complete to the level of lap time simulation input, estimated mass properties, and possibly estimated cost, labor, and other considerations. Each should be verified with a rough general arrangement. The generation of variants is more efficient if, once objective goals have been developed for acceleration in the axial, braking, lateral, and yaw axes, the performance-related properties of each conceptual design variant are tuned to these values. A good amount of work is necessary to develop the variants and the evaluation tools.
It might be convenient to start the development of conceptual design variants with the team’s last best racecar, since complete information on this design and its performance is available. But a central tenet of conceptual design is that most design problems have many different feasible solutions, and the optimal solution can be difficult to immediately identify. Once their conceptual design process is up and running, teams should take the opportunity to evaluate a wide range of very different (but feasible) FSAE racecar concepts in order not to miss the design concept that will most surely reward them with success.
Completion of the conceptual design process allows generation of system design requirements for the following embodiment design process. Thrust and resistance levels are available as design requirements for major systems. System and part masses and costs may also be passed on as embodiment design requirements (to be iterated back to the conceptual level, if found to be too easy or too difficult to achieve). The performance of the whole vehicle is then more likely to rise to the level of the sum of the system performances.
After sufficient examination of the FSAE racecar design space, a team should have both a proven best design concept (for their specific constraints) and a profound understanding of why they can rightly use that word – optimum.