+ Reply to Thread
Page 2 of 4 FirstFirst 1 2 3 4 LastLast
Results 11 to 20 of 36

Thread: Suspension Geometry Analysis

  1. #11
    I second what Kevin said, simple sketches in a CAD program is a great way to go although specific programs do give a wealth of very useful data and some very cool visualisations they tend to be limited in their applications, for example, neither susprog or OptimumK can simulate a front beamarrangement which uses a peg and slot although I will admit this is a very specific case.
    Aston University Formula Student - VD/Suspension guy.

  2. #12
    Junior Member
    Join Date
    Jan 2014
    Location
    Coimbatore,India
    Posts
    9
    Mr.Kevin

    Thank you very much for sharing your detailed insight into the problem. I found it useful.

    Regards
    Koushik

  3. #13
    Junior Member
    Join Date
    Jan 2014
    Location
    Coimbatore,India
    Posts
    9
    Christian

    Thank you too for stopping by the thread and seconding Kev's post.

    Regards
    Koushik

  4. #14
    To add to Kevin's post about the power of 3D CAD,

    I've used Solidworks Motion Analysis add-in to generate all of my kinematic data. You simply cycle the linkage in the motion study with the equivalence of a linear actuator in SW and use trace paths, angle projections, and other data recording add-ins. These can be exported directly to Excel to further interpret the data with graphs and any other mathematic operations needed.

    JR

  5. #15
    Koushik and all,

    Here is a bit of food for thoughts about how kinematics software is or could be working.

    At least a beginning of food for thoughts.

    Let’s take the generic case of a one 5 links suspension.

    You have 5 links. Each of them having an outboard pick up point O1, O2, O3, O4 and O5 and an inboard pickup point I1, I2, I3, I4 and I5 and a given length L1, L2, L3, L4 and L5

    Also there is a wheel center C and a tire contact patch T. In this example the wheel has an infinitely small thickness so the contact patch is only one point

    In this simplified example there is no compliance and there is no tire deflection.

    If the outboard point O1 moves it will be moving on a sphere which has I1 as a center and L1 as a radius. Similarly for all other 4 points.

    However there are constraints. Many of them. As we assume there is no compliance the distance O1 to O2, O2 to O3, well all the distance between all the outboard points … and also the distance between each outboard point On and C …. and also the distance between each outboard point On and T … cannot change.

    Therefore when the wheel center C goes up (vertically) by let’s say z mm there will be an imposed trajectory of all the O points as well as the point T. There will be one and only one track variation, wheelbase variation, camber, toe, caster, KPI variation…

    (Ouch… How do we measure the Caster and KPI angle on one 5 links suspension geometry? Let’s leave that for later and let’s simply say that there will be a rotation of the upright around the x axis and that will be the KPI and Camber variation and a rotation of the upright around the y axis and that will be the caster variation)

    Now take 2 points of the wheel / upright assembly as far away as possible from each other. Let’s take T and another point. Let’s say that this other point is O1 and that O1 is the highest outboard point.

    For z mm of the wheel center C vertical movement there will be a given movement in the x, y and z direction of point T and O1.

    Let’s call Ti the initial point of T and Tf the final point of T after the wheel center z vertical movement and similarly O1i the initial point of O and O1f the final point of O after the wheel center z vertical movement.

    Look at the segment Ti-Tf. Find the middle of this segment Ti-Tf (i for initial, f for final) Draw in your mind an imaginary 3 D plane perpendicular to the segment Ti-Tf in the middle of segment Ti-Tf. Let’s call this plane A.

    Similarly, look at the segment O1i-O1f. Find the middle of this segment O1i-O1f. Draw in your mind an imaginary 3 D plane perpendicular to the segment O1i-O1f in the middle of segment O1i-O1f. Let’s call this plane B.

    The intersection of planes A and B is your instant axis of the wheel about the chassis.

    Obviously you will have different instant axis position for different x axis movement so ideally you want a lot of small iterations of the wheel center z movement.

    If you want to know more have a look at http://papers.sae.org/2008-01-2949 New Method to Evaluate Bump Steer and Steering Influence on Kinematic Roll and Pitch Axes for All Independent Suspension Types

    I hope this helps.

    It took us several months to first create OptimumKinematics several years ago (and in fact Kevin Hayward was part of this effort when he was an intern with us) and a few other months a few years later to improve it. In fact we never stop improving it and there will be another version with forces and optimization probably next year (it could come in 2 months if we did not have any other consulting work to do…).

    As a design judge I would give you a lot of credit if you would be a FSAE / FS participant and you would come with your own kinematics software…..providing the time you dedicate to this software did not hamper a good job on the design and manufacturing of all other suspension parts and the car development / tuning. Your personal goal may not be your team goal.

    A few years ago I visited a FS team in Germany who wanted to design their own accelerometers. Good challenge but I asked them if they wanted to compete in Formula Student competition or in a Formula Accelerometer competition.

    The question is: why do you want to reinvent the wheel if you can buy it cheaper an quicker somewhere else? Only you can answer this question.

    If you want to do it, you now have one beginning of the answer at the beginning of this thread (I say one beginning because there could be other methods)
    Claude Rouelle
    OptimumG president
    Vehicle Dynamics & Race Car Engineering
    Training / Consulting / Simulation Software
    FS & FSAE design judge USA / Canada / UK / Germany / Spain / Italy / China / Brazil / Australia
    [url]www.optimumg.com[/u

  6. #16
    Senior Member
    Join Date
    Sep 2002
    Location
    Perth, Western Australia
    Posts
    717
    Koushik and all,

    Now that I have suggested that you continue developing your program I want to make the point that you should also use OpimumK (or an equivalent). The ECU team still uses OptimumK as part of its analysis procedure.

    Up front I will say that I was involved with the development of the first version of OptimumK. It was an amazing team of people to work with and I am very proud of what we achieved. I am probably a little biased when I say that I think it is the best suspension kinematics analysis software out there. What it does it does very well, and is very easy to use. I will also add that when I went to work at OptimumG I had already created a geometry analysis program over a period of 2-3 years. It used a different calculation method and none of it was useful for the creation of OptimumK, which was built for entirely different reasons. I found the two tools (my own and OptimumK) to be complementary rather than any one being a replacement for the other. You should think about what your approach can add to your workflow. Reinventing the wheel is not necessary, but there are lots of different types of wheels out there that are operated under different conditions.

    I have mentioned previously that OptimumK is not an ideal creation or iteration tool. It is difficult to force a wide user base into a particular method of geometry creation, and very likely to create more frustration than a better workflow. To that end everything is entered as x-y-z coordinates. You are probably already aware that is not the best way to approach suspension design. To alleviate this OptimumK can take a spreadsheet input so that geometry creation can be handled by something else. Excel is easily good enough for this.

    It is also not an optimisation tool. The main issue here is that suspension geometry by its nature is a multi-objective problem, which optimisation routines do not play well with. Furthermore optimisation of a purely kinematics problem is prone to lead to bad car design as the problem is much more complex. The only real way to build a decent optimisation tool for racing is to use lap-time as an objective function, and it takes one heck of a program to include a full kinematics solution inside of a decent lap time simulation.

    What it primarily serves as is an analysis tool. The main inspiration for the first OptimumK came from the easy to learn Solidworks UI and the Motec i2 software. The complete output data of a kinematics solve is to be able to be analysed and graphed as the user wants. Any input (with the assumption of a flat road) can be investigated. Comparisons between suspension systems is very easy.

    What is not apparent to the user is how adaptable the underlying code is. The heart of the software is a general purpose kinematics solver. This means that updates for OptimumG are quite straight-forward. If the company wanted to it could make the same software analyse just about any kinematic linkage (not just suspensions). The layer that sits on top of that that deals with different types of suspensions makes the whole thing easy to use. Are you aware that some commonly used suspension systems are either always over constrained, or in certain positions over-constrained? A good example is the old v8 supercar rear suspension with a watts linkage. There is a layer of code on top of the general purpose solver that deals with these issues, while keeping it easy to use. In practice I would say nearly all relevant suspension systems are covered by the software.

    I support the creation of your own software because I think one of the most important words for students in FSAE should be "and". Use your own software and SimMechanics and OptimumG. Build a good car and develop your program. FSAE students already stand out against other engineering students. But how do you stand out against all of the of the other FSAE students?

    Additionally whatever program you create will not be as polished and user friendly as OptimumK. This means it will likely be much less useful to others on your team. For the price of OptimumK I would imagine that every FSAE team should have it, and then develop their own tools to complement it.

    Kev

  7. #17
    Kevin,

    Thank you for your good words and the positive criticisms (with which I mostly) agree of OptimumKinematics. Also good memories from about.... gee.... 9, 10 years ago?

    A. I do agree with you about students making their own software: it helps them to better understand what they are dealing with... but making the software USEFUL with good graphical interface and user friendly features is another story: do they have the time? I have seen so many students Matlab simulation where you need to go in the codes to change the spring, you cannot overlay, you cannot zoom etc..

    B. Lap time.... Ah Lap time.....

    "The only real way to build a decent optimization tool for racing is to use lap-time as an objective function,"

    Lap time, yes but also
    - tire use (in % of use of each traction ellipse if possible at each meter of the track, traction ellipses shapes and size which never stop changing either)
    - balance (which in simple terms means do you have the yaw moment you want when you want)
    - consistency (which in simple terms means tire grip consistency with tire wear and temperature, track conditions and fuel load - that one not important on a FSAE car)
    - off and on center control (yaw moment generates by degree of steering (delta) for a given chassis slip angle (beta), this chassis slip angle being zero at the corner entry and "something else" at the apex
    - off and on center stability (yaw moment generates by degree of chassis slip angle (beta) for a given degree of steering (delta), this steering angle being zero at the corner entry and "something else" at the apex

    Control and Stability can be translated in an other words: drive-ability. We have found that lap time is actually a smaller factor than balance (which is yaw moment) control and stability. The reason is that these last 3 factors link the car with the driver with some quantification of the human-machine interface.

    In professional driver debriefing we discover that after a while there is clear link between o
    Om one hand
    - driver subjective data (Grip, control and stability are quoted from 0 to 10, 0 is bad 10 is excellent- but good drivers are never happy so I rarely see a 10 in their debriefing sheet - and balance quoted for - 10 being bad understeer and + 10 being bad oversteer and 0 being neutral)
    Amd on the other hand
    - and the control (Nm/ deg of delta for a given beta)
    - and the stability (Nm/ deg of beta for a given delta)

    When you have played with these other factors you will see that lap time is not as useful as a quantification of the car performance, Useful but not the # 1 parameters to consider. the best prove is setup a will suit driver 1 and setup B will suit driver 2 with equal lap time....

    The whole yaw moment Vs lateral acceleration graph is well explained in the Milliken book.

    C Kinematics Optimization; we are working on that (as well as suspension elements forces for given tire force sand moments input) be patient. That being said you can still use the automated batch run. Many students are not even aware of this function. That is because t do not read the help file but there little we can do for them; it is not a not an OptimumKinematics specific issue....

    D. Finally let's not forget that any pure kinematics software has its limitation because it doesn't take into account tire deflection and compliance. It is just a tool to compare camber, track, toe, caster, caster trail, scrub radius, etc... variations, roll and pitch axis position and movement (here we go again) for different kinematics solutions and different heave, roll pitch and steering movement (and/or combination of those).

    A DTM race car manufacturer bought 8 OptimumKinematics professional licenses. They just want to know, for example, if and how much bump steer there is when the chassis goes down 50 mm. Without tire deflection and/or compliance. THEN they use Adams. And they save a lot of time by doing so instead of inputting all kinematics and compliance.

    A passenger car manufacturer (you know the car with bushing at each suspension element end) does the same and also saves a lot of time.
    Claude Rouelle
    OptimumG president
    Vehicle Dynamics & Race Car Engineering
    Training / Consulting / Simulation Software
    FS & FSAE design judge USA / Canada / UK / Germany / Spain / Italy / China / Brazil / Australia
    [url]www.optimumg.com[/u

  8. #18
    Koushik,

    if you are interested in doing your own kinematics solver have a look on this website to get some idea's www.dynatune-xl.com. The tool is completely vector algebraic
    and can therefor calculate all types of independent (multi-link) suspension, do compliance analysis and predict accurately link loads, all in one go and in excel.

    Cheers,
    dynatune

  9. #19
    Senior Member
    Join Date
    Sep 2002
    Location
    Perth, Western Australia
    Posts
    717
    Claude (and others),

    Coming up on 10 years soon. A lot has happened in that time, but working with the group of people at OptimumG will always be a fond memory. Despite the stressful deadlines The big bonus was I learnt a lot of dirty French words, while I am still waiting for the opportunity to use this in life I am sure it will come in use one day.

    I definitely agree with the points you made about the usefulness of software. It is important to be very clear about what you want the software to do, and who it is for.

    The point I was making about lap-time is that just about any approach to optimisation I have seen requires single objective functions. Most multi-objective ones end up essentially as weightings on the various objectives to make one objective. Otherwise the outputs can be represented as points against different objective scales such as pareto sets. Lap time is not the only useful metric, but probably the one that is best suited to optimisation.

    It might be good to include things like driver/tyre condition as a set of random variables. This will help move the global optimisation towards solutions that are not overly sensitive. It is worth noting that while the performance of a car is not 1-1 with most variables the solution space is reasonably clean (i.e. not a lot of local minima). A decent heuristic optimisation routine should be able to deal with the random inputs and the solution space quite readily.

    There are ways to build good driver models into lap time sims, but the computation time starts increasing to a point where wrapping an optimisation routine around it starts taking a lot of computation time. This starts to get outside of the realms of usefulness for the race engineer at the average computer and into the more professional teams. In other words difficult to make money on. You only have to look at the large number of variables to get a concept of how much computation could be needed. It is a similar problem in other simulation fields. Shape optimisation is a good example. A single FEA run might take a few seconds. Multiply it by 10000 and you have a big portion of the day gone. 10000 sounds like a lot, but what if you had 10 variables each with 10 options only (i.e quite coarse). Your 10000 runs only cover 0.0001% of the solution space. Project this into multi-objective optimisation and the limiting factor quickly becomes the running time of your simulation. The main solution is to get it to run quicker, but that usually comes at a loss of fidelity. In the end you are searching through an abstracted solution space anyway so great care needs to be taken.

    I normally tell my students that this is a good thing otherwise all of us engineers would be out of a job pretty quickly if problems weren't this hard.

    For kinematics sims I think the batch runs are good enough. Optimisation brings its own problems, that can work against the user. What function do you optimise for? What variables are kept free? All of this working in such a neat user interface would be a big challenge. I am personally more excited about seeing the springs etc. going into the software. Also maybe starting to address the movement of the contact patch centre. I have this already in my own code, but the ease of use in a developed package will make for a lot less headaches. As always I bet the challenge for OptimumG is not in doing the math, but putting it into something foolproof with a decent UI. Once it is in the product easily goes into its own unique space. Somewhere smack between full VD capable software and the purely kinematics software, while still maintaining the ability to do pure kinematics. At a decent price point it would make the latter almost completely without a market.

    However my main preferences are for well designed tools that do specific jobs well, rather than all purpose tools that need a lot of fiddling around to get results. I am not surprised by the DTM manufacturer example. That would come pretty close to how I would work on the problem. Kinematics only for the simple things, iterate until you get something that looks reasonable and then go to more depth if necessary.

    Ultimately finding good solutions means exploring as much of the solution space as possible. Doing some initial searching with coarse tools allows you to narrow down to areas of interest, which you can then analyse in more depth. This would be akin to using a telescope before the microscope. If you only have a microscope you get a really good view of your immediate surroundings, but no clue as to what is only a small distance away.

    Kev

  10. #20
    Senior Member
    Join Date
    Mar 2005
    Location
    Modena, Italy
    Posts
    363
    I also made an excel based kinematic solver which also calculated link forces and I learnt a great deal. I was without adams for a few weeks and needed to get started on a 5 link suspension design. As a learning tool I can highly recommend it.

    The kinematics were an open form and iterative.

    The basic equations came from a vector analysis of the system where a vertical velocity vector of unit length is applied to the contact patch and the remaining 5 degrees of freedom are solved based on the kinematic constraints:
    1. The velocity vector of each point on the upright is a function of the 5 "response" degrees of freedom (dTrack, dWheelbase, dToe, dSpin, dCamber)
    2. The dot product of the axes of each link and the velocity vector of its outboard point = 0 (since they must be perpendicular)

    This gives you 5 equations and 5 unknowns which is easy to solve in excel in matrix form.

    Then you can do cool stuff like this:
    Roll centre height = Track x (dTrack / dZ) / 2
    Anti-dive/raise angle (braking) = dWheelbase/dZ
    Anti-lift/squat angle (acceleration) = some function of dWheelbase, dSpin & dZ but I don't remember now.
    Camber gain = dCamber
    Bumpsteer = dToe

    But the even cooler stuff is when you realise you can rewrite the equations so you input design parameters (roll centre height, anti angles, bumpsteer, and camber gain) and the equations calculate the pickup points.

    Its the duck's nuts.

    I use this method in parallel with my adams work as it allows you to change (for example) the roll centre height without touching the bumpsteer, even with a complicated 5 link setup. Otherwise you find yourself manually iterating forever in adams or optimumk to arrive at a set of pickup points that satisfy all of your design targets. In fact I'd even go as far as saying if you don't have SOME sort of (possibly simplified) support calculations going on beside your adams/optimum K model you aren't really engineering, you are just playing.
    Last edited by Tim.Wright; 03-11-2015 at 06:20 AM.

+ Reply to Thread
Page 2 of 4 FirstFirst 1 2 3 4 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts