You are on page 1of 30

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Design of Mountain Bike Hydraulic Disc Brake System (HDBS) Uncertainty Analysis

ME6105 Modeling and Simulation in Design Spring 2008

Prepared by: Derrick Ma Ethan Coffey Phil Grzemkowski

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

1. Evolution of the mountain bike hydraulic disc brake system Dymola model In the previous assignment, our group had developed an energy-based system model of the mountain bike hydraulic disc brake system (HDBS) in the Modelica language. In order to make the model and simulation of said model more manageable, it was determined that it was necessary to make some assumptions. Also, to avoid potential pitfalls and debugging issues, the overall system was decomposed into its subsystems, specifically: 1. The brake lever (master cylinder only) 2. brake caliper 3. brake pads (modeled via the standard braking component found in Modelicas rotational mechanics library) 4. rotors Using the bottom-up model building philosophy, each subsystem was further simplified and reduced to its core components as shown in Figure 1-1a to 1-1d. This approach allow the team to test and verify each subsystem is working as expected before they are combined to form the overall braking system.
flange_b1

Fixed2

cylinder1i

positionSensor

port_A1

Figure 1-1a. Dymola model of the brake lever subsystem (master cylinder)
f

f1
spring1 Fixed2
cylinder1i

port_A1

positionSensor2
s

Figure 1-1b. Dymola model of the brake caliper subsystem

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


inertia flange_a

J=100

Figure 1-1c. Dymola model of the rotor


y

pMeas x

mC

caliper2...

rigidLine

rotor1 qMeas ramp torqueSensor brake_pad tau fixed=0 duration=0.5

Force1

Figure 1-1d. Dymola model of the entire braking system.

It was discovered that the braking model (shown in Figure 1d) did not accurately predict the response of the braking system. This inaccuracy was attributed to the over-simplification of the braking system components. For instance, to better replicate the physical actuation of the brake caliper subsystem, an elastogap should be included to simulate the small air gap between the brake pad and the rotor when the braking system is at rest. Also, it was determined that the inertia of the bicycle may have been erroneously estimated. Therefore, prior to performing the uncertainty analyses, the Dymola model was improved upon by fine tuning each subsystems of the HDBS. Moreover, other detail such as modeling of the bicycle wheel and the road surface was added to the improved Dymola model as shown in Figure 1-2 below:

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


const

ramp k=0 const1 duration=0.5 y k=0 r...

world1

Brake_Torque...

tau

Caliper_inlet_Pres

b...

cal...

bike_inertia

speedSensor

qMeas

brake_pad fixed=0 gain realToBoolean timer

road

B
k=-1

Figure 1-2. Dymola model of the entire braking system (improved version)

It was determined that the enhancements made to the HDBS model had significantly improved the accuracy of the system response, particularly in prediction of the stopping distance and the stopping time (time required for a bicycle in motion to come to a complete stop)

rigidLine

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

2. Uncertainty analysis identifying and modeling of the uncertainties associated with the energy-based system modeling When deciding on the variables to be included in the uncertainty analysis of the mountain bike HDBS model, it was first decided to include most variables in the system. The purpose and rationale behind this decision was to gain a better understanding of the effects of each variable on the stability of the system; as well as to identify the design variable(s) that contribute significantly to the uncertainty of the measurement of effectiveness. The probabilistic analysis was performed multiple times, with varying number of runs. The rationale for performing the probabilistic analysis with various numbers of runs was to determine if there is any significant correlation(s) between the effects of probabilistic analysis on determining and isolating the key uncertain variables. The number of runs, n, varied between 20, 100, 500, 1000. It was decided that performing the probabilistic analysis over a minimum of 1000 runs (n = 1000), since the high running average as a result of performing 1000 runs is likely to damp out any outlying / strange effects of the ModelCenter simulations. As shown in Figure 2-1 below, it was determined that many of the variables of our system did not have a significant effect on the system with the exception of 6 variables.

Figure 2-1a. Central Composite Analysis with many variables (n = 1000)

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Figure 2-1b. Central Composite Analysis braking distance response histogram with many variables (n = 1000) After running the simulation with most variables it was decided that it is logical to focus on a new simulation where only the only the variables that had more of an effect on the system should be tested for the overall uncertainty of the HDBS model. The rationale behind this logical decision was based (partly) on intuition, since uncertainties or variability of certain variables will have a greater impact on the overall outcome of the system response. For instance, the variability of the friction (resistance) between the master cylinder piston against the master cylinder housing will affect the velocity of the master cylinder piston, thereby affecting the energy transfer efficiency from the brake lever into the hydraulic system. However, when compared to the affects of variability of the friction coefficient of the brake pad material, the overall affect of the resistance in the piston movement as a result of friction between the piston and the housing assembly is likely to become immaterial; and as such, the uncertainty of friction forces internal to the brake lever piston/housing assembly can be ignored and excluded from the uncertainty analysis. This aforementioned uncertainty variables elimination process in which our group had employed to deduce the applicable uncertainty variable is the essence of the elicitation method. This was accomplished by establishing a series of elicitation questions against all of the uncertainty variables; and from which 6 variables were

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

selected based upon what the group had deemed to be the most crucial. The variables that showed a greater effect were as follows with reasoning for the selection of limits on each. For clarity, the triangular distributions for each of the following uncertainty variables were summarized in Table 2-1. inputForce 50 to 200 N force on the brake lever applied by the rider with a center of 150 N of force. This force comes from information on grip force for average people1. rotorSpeed This is the angular velocity of the wheel. The start point of the design was an equivalent of 20 mph. The upper and lower bounds of 3000 and 100 rad/sec equate to a speed of 40 and 1 mph respectively. The reason for these limits is to understand the affects of a bike that is barely moving and also a bike that is really out of control traveling downhill. The assumption is that this rider really will depend on the brakes. frictionCoeff The friction coefficient of the brake pad against the rotor. This variable is the source of most of the design uncertainty. It is known that the value should be between 0.15 and 0.8 however the exact value in unknown. For the analysis a center of .45 was used. brakeGeometry The brake geometry is a function of the size of the rotor. The initial values that were tested are 0.0001 mm2 to 0.0003 mm2 with a center of 0.00028 mm2. These values are calculated using a 160 mm diameter rotor with a 110 plate ID. The value of 0.0001 mm2 equates to covering approximately 4 degrees of the rotor surface. This was assumed to be adequate, however after further investigation this would mean that the brake pad length at the outside diameter of the rotor would be only 5.5 mm. This is an unrealistic value for the size of the brake pad. The value of 0.0003 mm2 for the upper bound equates to a pad size of 14 mm at the rotor OD. This value makes a better lower bound for our designed brake pad. A new upper bound of 0.0006 mm2 is a better fit the upper limit of our test. caliperPistonArea This is the cross sectional area of the piston in the caliper that is pressing on the brake pad. Initial values of 5x10-5 to 15x10-5 m2 were used as limits for the analysis. These values also correspond to a piston diameter of between 8 and 14 mm respectively. This would be an unusually small caliper piston. New values for the piston diameter of 15 to 27 mm were used and the analysis was recalculated. inputForcePos The location on the brake lever where the force is applied. This is less of a design variable and more of a real uncertain input. As the rider is on the bike the location of his/her hand will change with time. This change can have

Edgren, C.S., Radwin, R.G. & Irwin, C.B. (2003). Grip Force Vectors for Varying Handle Diameters and Hand Sizes. University of Wisconsin, Madison, Wisconsin.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

a large impact on the performance of the brakes. The limits for the initial analysis were 20 to 120 mm. This corresponds to the placement of the riders hand from the center of rotation of the brake lever. Table 2-1 Triangular Distribution functions of the Uncertainty Variables Variable Lower Limit Upper Limit Center Value Input force 50 N 200 N 150 N Rotor (disc) Speed 750 rad/sec 3000 rad/sec 1500 rad/sec Brake pad friction coefficient 0.80 0.45 0.15 2 2 Brake geometry 0.0003 m 0.0006 m 0.00045 m2 Caliper piston geometry 0.00017 m2 0.00057 m2 0.00045 m2 Force application 0.020 m 0.120 m 0.075 m (Input force position) In addition, the set of data points from elicitation questioning process were inputted into a comprehensive cumulative distribution function (CDF) calculator to predict the probabilistic distribution function and inverse CDF values associated with each of the 6 uncertainty variables. The inverse CDF ranges are shown below in Table 2-2. Table 2-2 Inverse CDF (5th & 95th percentile) of the Uncertainty Variables 5th Percentile 95th Percentile Units Input Force 169.7337 117.6161 N Rotor Speed 1224.806 1700.000 rad/sec Friction Coefficient 0.2952096 0.7510998 Brake Geometry 0.0003707 0.0005556 m2 Caliper piston area 0.0002770 0.0004910 m2 Input Force Pos. 0.0330541 0.1000000 m Likewise, the CDF splines and the probability distribution function (PDF) for each of the 6 uncertainty variables are shown below in Figure 2-2a through Figure 2-2f:

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


Cummulative Distribution Function
1.0 0.9 0.8 0.7 Probability 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -250

-200

-150 X values

-100

-50

(Note: You can select the blue dots and drag them around to modify the li )

Figure 2-2a. CDF Spline Input force variable

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


CDF - rotorSpeed
1.0 0.9 0.8 0.7 Probability 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0

500

1000

1500

2000

2500

3000

3500

X values (Note: You can select the blue dots and drag them around to modify the li )

Figure 2-2b. CDF Spline Rotor Speed

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


CDF - frictionCoeff
1.0 0.9 0.8 0.7 Probability 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0

0.2

0.4 X values

0.6

0.8

(Note: You can select the blue dots and drag them around to modify the li )

Figure 2-2c. CDFSpline Brake pad coefficient of friction

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


CDF - brakeGeometry
1.0 0.9 0.8 0.7 Probability 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 X values (Note: You can select the blue dots and drag them around to modify the li )

Figure 2-2d. CDF Spline Variations in possible brake geometry

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


CDF - caliperPistonarea
1.0 0.9 0.8 0.7 Probability 0.6 0.5 0.4 0.3 0.2 0.1 0 0.0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007

X values (Note: You can select the blue dots and drag them around to modify the li )

Figure 2-2e. CDF Spline Variations in possible caliper piston geometry

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


CDF - inputForcePosition
1.0 0.9 0.8 0.7 Probability 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 X values (Note: You can select the blue dots and drag them around to modify the li )

Figure 2-2f. CDF Spline - Force application (Input force position)

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

3. Determining the Expectation and Standard Deviation of the Model Output With the uncertainty variables and CDF splines described above, a Central Composite Experiment (CCE) was performed to determine the main effects of these uncertainty variables on the overall responses of the HDBS model. The overall performance of the HDBS can be quantified or measured with the following 2 key responses: Braking time (time X required for the bike to come to a complete stop) Braking distance (at the onset of braking, time = 0, the bike position = 0, at the onset of complete stop, time and bike position = X)

The result of the main effect analysis of the Central Composite Experiment for stopping distance is shown below in Figure 3-1a. Based on resulting main effect plots, the uncertainty variable that may have the greatest, potential impact on the stopping distance response of the HDBS system is the rotor speed and the caliper piston area. Likewise the result of the main effect analysis of the Central Composite Experiment for stopping time is shown below in Figure 3-1b. Based on resulting main effect plots, the uncertainty variable that may have the greatest, potential impact on the stopping time response of the HDBS system is the rotor speed and coefficient of friction of the brake pad material. The result of these CCE main effect analysis are logical since the time required for a bicycle to come to complete stop and the accompanying braking distance is highly dependent on the speed of bike at the onset of braking, the amount of clamping force generated by the caliper piston as a function of the pistons geometry, as well as the braking torque generated as a result of friction between the brake pad and the rotor. It is obvious that the faster the bicycle is moving at the onset of braking will take a longer time to stop and a greater braking distance. Likewise, the same can be said of the friction coefficient; since excessive wear and tear on the brake pads or negligence in brake pad maintenance greatly affects braking performance, in particular since all the generation of braking torque is largely dependent on the available friction from the pads. If the pads were completely worn (smooth), the amount of normal force applied to the pads to clamp down on the rotor (also assumed to be smooth) must be greatly amplified.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Figure 3-1a. Central Composite Experiment (Mont-Carlo technique) - Main Effects Analysis of HDBS on stopping distance.

Figure 3-1b. Central Composite Experiment (Mont-Carlo technique) - Main Effects Analysis of HDBS on stopping time. To further investigate the effects of uncertainty variables and their respective impacts on the performance of the HDBS, the Monte-Carlo technique of

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

uncertainties characterization was used. Monte-Carlo simulation is a method for iteratively evaluating a deterministic model using sets of random numbers as inputs. This method is often used when the model is complex, nonlinear, or involves more than just a couple uncertain parameters. In this case, the ModelCenter software randomly generates uncertainty values (based on the Monte-Carlo technique) as inputs to the HDBS model. The resulting histograms from the Monte-Carlo technique of characterizing uncertainties are shown below in Figure 3-2a and Figure 3-2b.

Figure 3-2a. Probablistic analysis (Monte-Carlo technique) histogram of stopping distance as a function of the 6 uncertainty variables

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Figure 3-2b. Probablistic analysis (Monte-Carlo technique) histogram of stopping time as a function of the 6 uncertainty variables The results of the Monte-Carlo simulation can be summarized in the following bulleted list: Likely stopping distance = approx. 9 m Likely stopping time = 1.3 seconds

Again, it must be re-iterated that the results shown in the Monte-Carlo simulation were based upon varying the uncertainty variables which includes rotor speed (i.e. initial speed of the bicycle). For the sake of comparison, the above analysis was repeated using the Latin Hypercube Sampling technique. The Latin Hypercube Sampling technique (LHS) is based on a distribution of plausible collections of parameter values, in this case the plausible values for all 6 uncertainty variables; from a multi-dimensional probability distribution. Likewise, the main effect plots as well as the resulting histograms are shown below in Figure 3-3a and 3-3b, and Figure 3-4a and Figure 3-4b respectively.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Figure 3-3a. Latin Hypercube Sampling technique - Main Effects Analysis of HDBS on stopping distance

Figure 3-3b. Latin Hypercube Sampling technique - Main Effects Analysis of HDBS on stopping time

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Figure 3-4a. Probablistic analysis (LHS technique) histogram of stopping distance as a function of the 6 uncertainty variables

Figure 3-4b. Probablistic analysis (LHS technique) histogram of stopping time as a function of the 6 uncertainty variables For comparison and error-checking, these numbers can be compared to the mean and standard deviation from the Monte Carlo simulation discussed previously.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

The numbers are almost exactly the same for the two simulations. This means that 1000 runs was enough for the Monte Carlo simulation to be accurate, since the Latin Hypercube simulation did not add any advantage with the same number of trials.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

4. Performing a Sensitivity Analysis Having understood and established the uncertainties of the braking model, it is also advantageous to determine the variable sensitivities associated with the HDBS. Sensitivity analysis is used to determine how sensitive a model is to changes in the value of the parameters of the model and to changes in the structure of the model. Sensitivity analysis is usually performed as a series of tests in which the modeler sets different parameter values to see how a change in the parameter causes a change in the dynamic behavior of the system response. Sensitivity tests help the modeler to understand dynamics of a system. By experimenting with a wide range of values, it can offer insights into behavior of a system in extreme situations. Moreover, discovering that the system behavior greatly changes for a change in a parameter value can identify a leverage point in the model. By showing how the model behavior responds to changes in parameter values, sensitivity analysis is a useful tool in model building as well as in model evaluation. Sensitivity analysis helps to build confidence in the model by studying the uncertainties that are often associated with parameters in models. Many parameters in system dynamics models represent quantities that are very difficult, or even impossible to measure to a great deal of accuracy in the real world. Therefore, when building a system dynamics model such as the HDBS system, the modeler is usually at least somewhat uncertain about the parameter values he chooses and must use estimates. Results from the sensitivity analysis allow the group to determine what level of accuracy is necessary for a parameter to make the model sufficiently useful and valid. For instance, if the tests reveal that the model is insensitive, then it may be possible to use an estimate rather than a value with greater precision. Sensitivity analysis can also indicate which parameter values are reasonable to use in the model. If the model behaves as expected from real world observations, it gives some indication that the parameter values accurately represent the real world. Based upon the 5th and 95th percentile figures from each uncertain variables inverse CDF from section 2, the group generated the MorrisExperiment.csv file with MATLAB (see Appendix A for more detail). The MorrisExperiment.csv file was inputed and linked into ModelCenter to generate MorrisResults.csv file. The MorrisResults.csv was fed into MATLAB to generate Morris plots (see Appendix B for more detail) as shown below in Figure 4-1 and Figure 4-2.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Figure 4-1. Morris plot for Braking Distance response

Figure 4-2. Morris plot for Braking time response

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

The elementary effects used in the Method of Morris screening for sensitivity analysis are as follows: Effect number 1: input force (from the rider) Effect number 2: rotor speed Effect number 3: friction coefficient Effect number 4: brake geometry Effect number 5: caliper piston area Effect number 6: input force position (position of riders hand on the brake lever)

It can be seen from the resulting Morris plot of the braking distance response (Figure 4-1), that effect number 2 and 3 (rotor speed and friction coefficient respective) has a large mean and large standard deviation of elementary effects; which suggests that the likelihood of these two factors having any effect on the braking distance is relatively high with respect to the factors 1, 4, 5 and 6 in other words, factors number 2 and 3 are suspects to have a significant interaction effect or higher order impact on the system response in terms of stopping distance. Furthermore, the Morris plot for braking distance response suggests that factors 1, 4, 5 and 6 (input force, brake geometry, caliper piston area, input force position respectively) also have an impact on the stopping distance but to a lesser degree compared to rotor speed and friction coefficient; though factor 6, input force position is likely to be immaterial in determining the performance of the HDBS system. The Morris plot result for braking distance was shown to be in agreement with the Monte-Carlo simulation for Main Effect Design of Experiment (DOE) analysis shown below (duplicate of Figure 3-1):

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Likewise, it can be seen from the resulting Morris plot of the time-to-stop response (Figure 4-2). As previously stated, the effect number 2 and 3 (rotor speed and friction coefficient) have a large mean and large standard deviation of elementary effects; which suggests that the likelihood of these two factors having any effect on the time-to-stop response is relatively high with respect to the factors 1, 4, 5 and 6 in order words, factors number 2 and 3 are suspects to have a significant interaction effect or higher order impact on the system response in terms of time required for the bicycle to come to a rest. Furthermore, the Morris plot also response suggests that factors 1, 4, 5, and 6 (input force, brake geometry, caliper piston area, input force position respectively) also have an impact on the time-to-stop but to a lesser degree compared to rotor speed and caliper piston area. It is also important to note that the Morris response plot for both the stopping distance and the time-to-stop response are identical from the standpoint of having the effect factors having the same impact on the system response may it be stopping distance or time-to-stop. This is logical and was as expected since braking distance and time-to-stop are linearly related. Therefore, to expect something otherwise would suggest a potential error in the HDBS Dymola model.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

5. Lessons Learned One of the most interesting lessons learned from this assignment is the main effects of the variables on the overall response of the HDBS model. Our group had expected a higher or more significant impact on braking performance being derived from having a larger contact area between the brake pad and the rotor. However, when this factor was compared with the 2 highly significant parameters, namely initial rotor speed (i.e. initial bicycle velocity at the onset of braking) and coefficient of friction of the brake pad, it is reasonable to see how contact area is relatively less impactful on the braking performance. Also, it was suspected that the geometry of the rotor, namely rotor diameter, would have a significant impact on braking performance. This is due to a larger braking torque armature resulting from a larger rotor diameter. However, the group did not include this as one of the uncertainty variables to be analyzed in the system, since rotor diameters are typical and common (implied standardization) between various bicycle components manufacturers. Therefore, such commonalities between component manufacturers effectively nullify the rotor diameter as an uncertainty variable particularly since a different rotor diameter would require a different caliper mounting design and location on the frame; consequently, typically the rotor diameters differ between the Forward and Aft (rear) rotors only as a function of brake bias, and not between various equipment manufacturers.

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Appendix A: MATLAB scripted used to generate MorrisExperiment.csv file


%This script generates a file called MorrisExperiment.csv which specifies %the runs for a Method of Morris screening experiment.

%SPECIFY FACTORS %Factor_01 = inputForce (min = 25 N, max = 150 N) %Factor_02 = rotorSpeed (min = 500 deg/sec (6.44 mph), max = 3000 deg/sec (38.67 mph)) %Factor_03 = frictionCoeff (min = 0.15, max = 0.65) %Factor_04 = brakeGeometry (min = 0.000277778 (2 deg), max = 0.000833333 (6 deg) %Factor_05 = caliperPistonArea (min = 0.0001266769 m^2 (piston dia = 0.5"), max = 0.004948315 m^2 (piston dia = 1.25")) %Factor_06 = inputForcePos (min = 0.02 m (0.78"), max = 0.12 m (4.72")) %NOTE: brakeGeometry = brakepad contact area % SPECIFY LOWER AND UPPER LIMITS xlb = [-169.7337, 1224.80, 0.295, 0.00037076, 0.000227, 0.03305]; %lower lim xub = [-117.6161, 1700.00, 0.751, 0.00055556, 0.000491, 0.10000]; %upper lim % SPECIFY NUMBER OF RANDOM OBSERVATIONS r = 40; % the number of random observations k = length(xlb); % the number of factors e = morris_experiment(k,r,xlb,xub); csvwrite('MorrisExperiment.csv',e)

function X = morris_experiment(k, r, xlb, xub, seed); % % % % % % % % Meaning of the variables: we use the same variable names as in the paper k = number of input factors p = grid_level (should be even) r = the number of effects that one wants to sample lb = optional lower bound on the x values ub = optional upper bound ont the x values seed = optional random number generator seed

m = k+1; % number of experiments per batch n = m*r; % total number of experiments % pick p to be something large so that it is unlikely that % the same grid point will be sampled twice p = r*10000; delta = p/(2*(p-1)); % check for lower and upper bounds if nargin < 4 xlb = zeros(1,k); xub = ones(1,k); end % seed the random number generator if nargin==5 rand('state',seed); else rand('state',sum(100*clock)); end % % % % % % J B %define sampling matrix of the form B = [0 0 0 0; 1 0 0 0; 1 1 0 0; 1 1 1 0; 1 1 1 1]; = ones(m,k); = tril(J,-1);

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


X = zeros(n,k); for i=1:r Dstar = diag(floor(rand(k,1)*2)*2-ones(k,1)); xstar = floor(rand(1,k)*p/2)/(p-1); Btemp = ones(m,1)*xstar+delta/2*((2*B-J)*Dstar+J); Bstar = Btemp(:,randperm(k)); X((i-1)*m+1:i*m,:) = ones(m,1)*xlb+ones(m,1)*(xub-xlb).*Bstar; end return;

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis

Appendix B: MATLAB scripted used to generate Morris plots


%This script generates Method of Morris plots to analyze the results of a %Method of Morris screening experiment. m = 2; % number of responses % E k e M load experiment and response data from MorrisResults.csv = csvread('MorrisResults.csv'); = size(E,2) - m; = E(:,1:k); = E(:,(k+1):(k+m));

%PLOT COMMANDS %The following lines contain plot commands to generate the morris plots. %There is one plot command for each response. Each command consists of %three lines of code. Add or remove plot commands as needed. Be sure to %index the effective mean and standard deviation variable names as well as %the column in the M vector for each additional response. Also, replace %the generic "Reponse #" with a more descriptive name. figure, [eff_mean_m1, eff_std_m1] = morris_plot(e,M(:,1)); title('Method of Morris - Response 1: Braking Distance') figure, [eff_mean_m2, eff_std_m2] = morris_plot(e,M(:,2)); title('Method of Morris - Response 2: Time until full stop')

%STATS %The effective means and standard deviations for each factor and each %response are stored in the stats matrix below. This is helpful for %determining which dot on the plot corresponds to which factor, because %the labels often overlap or are not sufficiently close to the dot. Notes %for interpreting the stats matrix are included below. Edit this command %to include the effective means and standard deviations for each response %that you are studying. stats = [eff_mean_m1', eff_std_m1', ... eff_mean_m2', eff_std_m2'] function [eff_mean, eff_std] = morris_plot(X, Y) % we assume that p large is and that therefore delta = 0.5; k = size(X,2); m = k+1; r = size(X,1)/m; % compute the elementary effects xdiff = diff(X); ydiff = diff(Y); % eliminate the meaningless datapoints xdiff(m:m:end,:) = []; ydiff(m:m:end,:) = []; for i=1:k ind = find(xdiff(:,i)~=0); eff = ydiff(ind)./(delta*sign(xdiff(ind,i))); eff_mean(i) = mean(eff); eff_std(i) = std(eff); end %plot elementary effects plot(eff_mean,eff_std,'o','MarkerEdgeColor','k','MarkerFaceColor','r','MarkerSize',8) hold on axisvals = axis; offset = (axisvals(2)-axisvals(1))*.06; line_vect = [axisvals(1),0,axisvals(2)];

Design of Mountain Bike Hydraulic Disc Brake System Uncertainty Analysis


plot(line_vect, abs(line_vect*sqrt(r)/2), '--b') axis(axisvals); for i=1:k text(eff_mean(i),eff_std(i)+offset,sprintf('%g',i),'FontSize',15); end; hold off grid on xlabel('Mean of Elementary Effects') ylabel('Standard Deviation of Elementary Effects') title('Method of Morris'); return;

You might also like