Modelling windscreen wipers and their eciency.

May 4, 2012
Abstract We previously used a given model to calculate the area covered by a single rear windscreen wiper. We used particular values for variables and parameters approximately matching those of two well known cars, the Vauxhall Vectra and the Vauxhall Agila. The previous approach to modellling the windscreen wiper did not actually attempt to maximise the area wiped and we shall now try to perform calculations in an eort to improve the eciency of the windscreen wiper in terms of the area it wipes.

Introduction

Windscreen wipers are essential to safety when driving a vehicle in rainy weather. They ensure good visibility and, of course, the larger the area of the windscreen they wipe, the better the visibility for the driver. We can model this area and the eectiveness of windscreen wipers using Mathematics and that is what we intend to do. In our previous model, we set out to look at the area wiped by a single rear windscreen wiper. It is clearer if we look at the diagram illustrating the situation:

Figure 1: The basic diagram of our windscreen and wiper. As seen in gure 1, we have several variables and parameters including the screen width b, screen height h, horizontal displacement of the wiper x, vertical displacement of the wiper d, length of wiper 1

arm L, length of wiper blade r, the angle of rotation and the angle between the wiper arm and blade . We make the assumptions that the vertical displacement d is equal to zero, the window is rectangular and that the wiper takes symmetric sweeps of /2. These will all factor into our calculations. In our previous simple model we simply investigated the resulting area that was wiped from given data and looked at this wiped area as a percentage of the total window area (giving us the wipers eciency). However, now we wish to in some way maximise the area and increase this percentage. How should this be done? Which variable should we maximise? We could choose any! We decided to maximise the angle in our improved model and in the process we set out to see how much we could increase wiper eciency of the two given car models as a result.

Data Collection

In the previous model we were given measurements for two Vauxhall car models (The Vectra and the Agila). We shall use mostly the same data in our improved model. However, some variables will be decided after we make certain calculations to decide the best values to use. Table 1: Measurements And Data. Symbol Vectra Agila b 100cm 107cm h 80cm 45cm 45cm 36cm r 30cm 23cm L ? ? ? ? 20cm 10cm x Later in the model we go on to change the value of x. However, for the rst part of our model we shall use the x values as displayed in the table above.

Mathematical Model

Figure 2: A diagram of the total area covered in general by the model. 2

The above diagram shows the movement of the windscren wiper and the area it covers (the shaded area). This diagram takes into account the area, but we must remember to account for the size of the window that will surround this sector. In the previous study, the following formulae were derived for the area A and the angle : A = ( )r(r + 2Lcos()) 2 and tan(/2) = L + rcos() rsin()

Rearranging the latter equation to get in terms of , and substituting this into the equation for (A) along with the values of (L) and (r) for the Vauxhall Vectra, we get : A = 45arctan( 1 30 + 45cos() )(45 + 60cos()) 45 sin()

Plotting a graph of the area in the regions that interest us (that is 0 /2), we see that the lower the value of , the higher the value of the area:

Figure 3: A graph of the total area wiped by teh Vectra against the value of angle . However, we cannot simply let = 0 and be done with it! Recall that earlier I mentioned that we must be wary of the size of the windscreen surrounding the sector that is wiped. In the previous study we derived two inequalities that prevented the wiper from overstepping the edges of the windscreen both horizontally and vertically. These were: Horizontally: Vertically: + rcos 2 Lcos + rsin 2 2 x + Lsin b + 2 2 2 + dh 2

Of course, we have set d = 0. Substituting in our formula for as well as values for x, L, r, b and h for the Vectra we get two substantially more complex inequalities shown on the following page:

Horizontal restriction: 30(30 + 45cos()) 20 + 45sin arctan (30 + 45cos())2 sin() 2025 + sin()2 Vertical restriction: 1350 (30 + 45cos())2 2025 + sin()2 + 45cos arctan

1 30 + 45cos() 45 sin()

50

1 30 + 45cos() 45 sin()

80

Now, solving these equations analytically would be a mighty task, so we can instead plot a graph of the left hand side of each inequality and nd out where the graphs lie below the respective right hand side of each inequality.

Figure 4: Respective graphs of the horizontal and vertical inequalities against for the Vectra. The dotted lines show the maximum values that the functions can have according to the inequalities and the blue regions below therefore show the feasable regions that will work when picking a value of . We know from gure 1 that we wish to have a value of as small as possible. Solving for where the horizontal restriction is equal to 50 (that is, where the graph is circled above), we get that f () = 50 when = 0.9148267775. Therefore we need 0.915 /2 for f () to be in the feasable region. Solving for the vertical restriction is easy. In fact for this function, f () is in the feasable region for all in the given range 0 /2. Therefore the only restriction that applies is the rst one. All in all, we have 0.915 /2 for f () and we would like as small as possible to cover the maximum area possible. This gives = 0.915. Having found for the Vectra we can now nd its respective using our formula: = 2arctan L + rcos() rsin() = 2arctan 4 30 + 45cos(0.915) 45sin(0.915) = 2.030

Following the exact same procedure but instead using the x, L, r, b and h for the Vauxhall Agila to nd its respective area equation and inequalities, we can deduce the best value of to use given those values. Furthermore, we can again nd for the Agila. Having done this, we end up with the following results for the Vectra and the Agila: Table 2: Derived values of and . Symbol Vectra 2.030 0.915 Agila 2.443 0.569

Results

We now have the data necessary to nd our new value of A, the area covered (by simply substituting our values into the area equation for each car model). The previous model judged the wipers performance on the percentage of the total windscreen area wiped (the eciency). Using all of our data and comparing with the results from the original model, we get the following results: Table 3: A comparison of the area covered and eciency of the original and improved models. Vectra Agila Old Area Covered 3328cm2 3125cm2 Old Eciency 42% 65% New Area Covered 3727cm2 3287cm2 New Eciency 47% 68%

As we can see from the above table, our maximisation technique has gone a little way to improving the eciency of the wiper on both car models. The values we have used for parameters and variables should now satisfy the inequalities that we took into account. As we can see, they are satised for the Vectra: Horizontally: 20 + 30sin 2.030 2 2.030 2 + 45cos 2.030 + 0.915 2 2 2.030 + 0.915 2 2 = 49.97 50

Vertically:

30cos

+ 45sin

= 60.60 80

They are also satised for the Agila: 2.443 2 2.443 2 2.443 + 0.569 2 2 2.443 + 0.569 2 2

Horizontally:

10 + 23sin

+ 36cos

= 53.47 53.5

Vertically:

23cos

+ 36sin

= 36.48 45

Therefore, we have an improved model that increases the eciency the wiper on each model we looked at by at least 3%. However, we should notice that after assuming that d (the vertical distance of the wiper) is equal to zero, we are in fact only wasting area by having any horizontal distance (x). Furthermore the model can be improved again by also assuming that the horizontal distance of the wiper x = 0. 5

So by again xing the values of L, r, b and h but also setting x = 0 as well as d = 0, we repeat the maximisation process we have used to plot the inequalities and nd a maximal value for given our parameters. After this we now derive the following angles for the Vectra and Agila: Table 4: Derived values of and . Symbol Vectra 2.349 0.656 Agila 2.735 0.332

This of course brings new results for the area covered and the eciency. These are very promising indeed: Table 5: A comparison of the area covered and eciency of the original and improved models. Vectra Agila Old Area Covered 3328cm2 3125cm2 Old Eciency 42% 65% New Area Covered 4891cm2 3913cm2 New Eciency 61% 81%

It can then be easily vered that the inequalities are satised as expected (without showing unnecessary working here again): Vectra: Agila: Horizontally we have 49.98 50 Horizontally we have 53.49 53.5 Vertically we have 50.67 80 Vertically we have 23.01 45

Finally, as a result of our new model, using the given L and r values to try and optimise the area wiped we have succeeded in achieving a 19% eciency increase for the Vectra and a 16% eciency increase for the Agila.

Discussion

I feel that our improved model signicantly betters the old one as it employs a logical procedure to select appropriate values for the angles involved in the wiper problem. However, the model makes this calculation based on the predetermined parameters L, r, x and d. For example, we could attempt to nd a way to maximise a function of L, r and , taking into account all three of these variables. This would then make the solution more dynamic as the combination of L, r and would be maximised, rather than just , (given a xed L and r). This would certainly bring even more eective results, as the values of L and r we were given in the original model are most likely not optimal. Certainly, the step where we made x=0 for both models made a huge improvement in eciency for both models, so it is clear to see that the other parameters do play a key role in wiper eciency. Therefore, our model improves the original one substantially, but we can condently say that the eciency of the wipers on our two car models could almost denitely be improved even more so using more advanced computations. Aside from changes between the original and new model, we should consider continuity. Both models make assumptions that do not reect the use of windscreen wipers in practice. For example, the vertical distance of the wiper would most likely not be level with the lower edge of the window (d = 0) and the wiper may not wipe with symmetric sweeps of /2. Most importantly, in my opinion, car windscreens are never rectangular but have a much more complicated shape, taking 6

on more of a three-dimensional one rather than the two-dimensional one we have modelled on both occasions. Indeed, these assumptions imminently result in the model being unrealistic in some senses and moreoever leaves it with what are perhaps large inaccuracies. Having said that, removing these assumptions adds unprecedented complexity to the problem. Without delving into such complexities, our model does what it sets out to do.

References
To be completed.....