Prototype for automotive roll over safety system

Prototype for automotive roll over safety

system

Ilias Tevetzidis

Automotive Engineering, bachelor's level 2020

Luleå University of Technology

PROTOTYPE FOR

AUTOMOTIVE ROLL OVER

SAFETY SYSTEM

Acknowledgments

Summary

This thesis presents the theory, methodology and results of building, tuning and testing a prototype for controlling body roll movement. Body roll is a result of lateral acceleration that acts on a vehicle when turning and result in torque on the roll center of the vehicle. The objective of this thesis is, in collaboration with Freno Air AB [7], to build a prototype and, with the help of sensors and micro controllers, restrict the body roll to avoid roll over on a semi-trailer that the company builds. The control unit in this thesis includes two PID controllers that is vastly used today in many different applications. The control unit is fed with signals from two pressure sensors on each side of the prototype frame. The pressures are given by four syringes, two on each side that are connected to the pressure sensors. One PID has small gains of KP = 0.2, KD = 0.005 and KI = 0 to take care of smaller and slower variations of pressure

difference and one PID with higher gains of KP = 1, KD = 1 and KI = 0 to take care of faster and

Tabel of Contents

Chapter 1: Introduction ... 1

1.1 Background ... 1

1.2 Semi-trailers history and how they work ... 3

1.3 Freno Air AB ... 5 1.3 Previous work ... 6 1.4 Objectives ... 6 1.5 Thesis boundaries ... 6 Chapter 2: Theory ... 7 2.1 Body roll ... 7 2.2 Control system ... 8 Chapter 3: Methodology ... 11 3.1 Experimental setup ... 11 3.2 Block diagram ... 16 Chapter 4: Result ... 19

Chapter 5: Conclusion & future work ... 21

1

Chapter 1: Introduction

1.1 Background

Ever since the introduction of embedded systems in automotive vehicles, it has been easier to develop features that would otherwise be impossible to implement. In parallel, stricter regulations came along that would challenge manufacturers to develop and manufacture more fuel efficient and safer vehicles.

It was not until the late 60’s that various federal, state and local governments in the United States linked a large portion of air pollution to the automobile after conducting studies on many different sources of air pollution. After that discovery, in 1967, the first agency was formed in the state of California focusing on air pollution. The agency was called “California Air Resources Board” [1] and three years later the federal “United States Environmental Protection Agency” (EPA) [2] was established to create and enforce emission regulations for automobiles. At the same time similar agencies were established in other parts of the world that enforced similar regulations.[3]

One of the first regulations to control air pollution by vehicles was the PCV (Positive Crankcase Ventilation) system. This system is still in use today and the way it works is that all the fumes and pollutive gases in the crankcase get ventilated and redirected to the intake of the engine to combust and burn rather than ventilating to the atmosphere [3].

Some of the first examples of fully electrical systems to be more environmentally friendly was the Electronic Fuel Injection (EFI) system that was developed by Bendix and one of the first cars it appeared on was Chrysler’s 300D in the late 50’s, see figure 1.1. Some of the benefits of the EFI systems was that it made the vehicle more fuel efficient, better throttle response, less maintenance etc.

2

Automotive safety was a concern from the very beginning of mechanized road vehicles. In fact, it was so dangerous that vehicle manufacturers implemented safety features in the vehicles without being enforced by government regulations. Some of the first examples are the four-wheel hydraulic brakes, first introduced in 1922 by Duesenburg Model A, safety glass that became standard in all ford cars in the 30’s etc. [4]

The first action by the American government was the “National Traffic and Motor vehicle Safety Act” that was enacted in 1966 to enable the federal government to set new safety standards for vehicles and road traffic safety. In the same year the U.S established the “United States Department of Transportation” (DOT) [5] with automotive safety as one of their biggest concerns.

Figure 1.2. Simple sketch of roll over event.

Today there are all kinds of supportive systems, some of them reduce pollution and others help and protect the driver actively while driving, or passively, when involved in an accident. One of those systems in focus, prevents a semi-trailer to roll over on its side while driving, see figure 1.2 and 1.3. This can happen if the driver turns too rapidly, while driving or if the semi-trailer is loaded poorly. Today there are some active safety systems that prevent roll-over for semi-trailers and some studies about the subject, but the majority of the systems are for commercial 18-wheelers, see figure 1.4, that require a professional driving license to drive.

3

One of those systems in production today is Wabco’s RSS (roll stability system) [8] that is used and installed in larger semi-trailers with air brakes. The module can be a replacement of the original ABS (Anti-lock Braking System) of the semi-trailer. Wabco’s RSS module comes with ABS integrated so the reqiurements and safety precations are kept intact.

Figure 1.2. 18-wheeler semi-trailer truck.

1.2 Semi-trailers history and how they work

In 1896 a man named Alexander Winton from Cleveland, Ohio started his business in manufacturing cars. One problem he came across was delivering the cars to the buyers. That wasn’t a problem for buyers that lived near, but for buyers across the county, that became a problem. Thoughts about driving the cars all the way to the buyers for delivery was fast dismissed because of expenses and the wear that would cause to the cars. That led him to invent a concept he called an automobile hauler which is essentially a semi-trailer truck to be able to load and transfer the cars that was manufactured and sold. He started manufacturing the haulers in 1899 for his own business but also sold the hauler to other automobile manufacturers. [9]

The semi-trailer truck is a very important tool for transporting goods nationally and internationally. Trucks transport over 70% of goods over land and there are 6,5 million trucks in circulation just in the European Union. Despite the increase in truck transportations the fatalities in truck accidents decreased with 53% from 2001 to 2014 [10]. Some of the semi-trailer manufacturers in Europe are:

• DAF Trucks [11] • Iveco [12] • Kamaz [13]

4 Figure 1.4. Basic inputs for semi-trailer.

The coupling between the semi-trailer and the tractor is called the “fifth-wheel coupling” which consist of a kingpin, vertical steel pin, protruding from the bottom front of the semi-trailer and the fifth wheel, see figure 1.5, a horseshoe-shaped device that the kingpin is inserted in. This coupling allows for all three rotations with some restriction to roll and pitch.

Figure 1.5. “Fifth-wheel” coupler.

The three basic inputs to the semi-trailer is air pressure for the air suspension and the air brakes, power supply to supply voltage to different electrical components such as the ABS unit and lighting, and finally the light and different function signals for the brake lights, turn indicators, axel lift etc.

When air pressure is not supplied to the semi-trailer the air brakes will engage for safety reasons, either if for some reason the air supply is not there or if the trailer is decoupled so that it stays in place.

5

1.3 Freno Air AB

Freno Air AB [7] is a company that develops and builds trailers and installs air suspension solutions for a variety of vehicles, all from pick-up trucks to trailers and vans, see figure 1.6. They have recently been asked to either develop and implement an anti-roll safety system in their semi-trailers or buy a similar system from other suppliers. This came as a request from another vehicle manufacturer that provides Freno Air AB with vehicles for their semi-trailers. This request is not a regulation, but more and more companies require this kind of system to increase safety.

Figure 1.6. Freno air AB semi-trailer with tractor.

The semi-trailers that Freno Air AB provides can be driven by anyone with a BA driving license. This particular driving license allows the towing of a trailer with maximum weight of 3500kg. Today most professional truck drivers get educated on how to distribute the weight of a load properly across the axels of the vehicle. For this particular application there is no extensive education on weight distribution when applying for the BA driving license and most common drivers do not think about it as much, thus this anti-roll safety system is required for safety.

6

1.3 Previous work

This thesis is a continuation of previous work in a previous course called Technical project 3. The previous works subject was to generate a concept, with all the intermediate steps in product development, for a roll prevention system that would be tested in computer simulations. The goal of the course was to learn about all the steps involving product development and the different type of methods used by companies to do so. The class was divided in three groups. One group was responsible for the CAD (Computer Aided Design) assembly of the semi-trailer to allow for co-simulations. One group was responsible for the dynamics of the semi-trailer and making relevant calculations. The last group, which I was a part of, was responsible for the software development and control system.

The project started by doing a benchmark, which is research in existing roll prevention systems, and reviewing customer feedback to make a list of specifications and customer needs that the concept ideas would be weighed against. The next step was to brainstorm ideas with no restrictions. Afterwards, all the brain storming ideas was collected, and they were weighted against the specifications and customer needs to find the best fitting concept idea. When the concept idea was determined each group made a Gantt-schedule with objectives to complete in a certain time frame and began working according to that schedule. The team managed to present a concept to Freno Air AB but was not successful in co-simulating the developed control system with the CAD files of the semi-trailer due to some technical difficulties.

1.4 Objectives

The main objective of this bachelor thesis is to build a small-scale prototype where an existing control system from previous work will be tested on, optimized and analyzed to determine if it is a viable option to further research and develop. The prototype will use one type of sensors that was determined in previous work, but all other sensors will be included in the methodology chapter. To start of the experiment one type of sensor will be used to see whether the system’s responsiveness is good enough or if it requires some additional information and additional sensors to be more accurate. Some other objectives that will be resolved in this thesis:

• Will the information of one type of sensor be enough to control and limit the roll movement?

• What type of information can the one type of sensor give about the state of the prototype?

1.5 Thesis boundaries

7

Chapter 2: Theory

In this chapter the theory of vehicle roll dynamics and control system will be presented and explain how they are connected to this experiment.

2.1 Body roll

Body roll is the result of centrifugal forces that are acting on a vehicles center of mass while cornering. In other words, it is the load transfer of the vehicles weight towards the outside of a corner. Vehicles with lower center of mass with respect to the roll center tend to roll less, see figure 2.1. That is why, vehicle manufacturers are pursuing for a lower center of gravity to improve the overall handling characteristics.

The torque occurring on the roll center (RC) is a result of the lateral acceleration (light blue arrow) times the distance between the center of gravity (CG) and the roll center (RC).

𝐹 = 𝛼𝑙𝑎𝑡∙ 𝑆𝑝𝑟𝑢𝑛𝑔 𝑚𝑎𝑠𝑠 (1)

𝜏𝐶𝐺 = 𝐹 ∙ 𝐵 (2)

1 in 2 yields,

𝜏𝐶𝐺 = (𝛼𝑙𝑎𝑡∙ 𝑆𝑝𝑟𝑢𝑛𝑔 𝑚𝑎𝑠𝑠) ∙ 𝐵 (3)

Figure 2.1. Depicts how body roll occurs.

8

2.2 Control system

A dynamic system is a system that is affected by external forces and disturbances over time. Dynamic systems are everywhere around us in the real world. Our bodies consist of many dynamic systems. These systems are often interconnected in such a way that the output of one system is the input to another system, this is called feedback. Feedback is necessary to know the state of another system. Otherwise, it would be impossible to control another system without knowing its condition first. For example, the pancreas in a human body takes the glucose level in the blood as an input and releases insulin or secretes glucagon as output, insulin and glucagon levels in turn becomes the input to the liver. In an event where we eat food, the glucose level in the blood rises which tells the pancreas to release insulin. The increase in insulin levels causes the liver to store excess glucose in the liver, thus keeping the glucose level in the blood constant. While the liver stores the glucose, the glucose level in the blood drops, which causes the pancreas to release less and less insulin until the glucose level in the blood has dropped to its target value [15].

Figure 2.2. Closed loop system

PID (proportional-integral-derivative) controller is the most common control system for utilizing feedback, see figure 2.2. It takes the feedback and subtracts it from the setpoint. Setpoint refers to the value that the user determines, for example you want a robot to walk 50 meters, the setpoint will then be 50.

9

Figure 2.3. Proportional part, error vs actuating signal with Kp of one.

Using the proportional part only can work in some situations but in some cases it will not because of the steady-state error. For example, if there is a need for a robot to walk 50 meters and stop, the proportional gain (Kp) is 0.1 and the actuating signal is in form of speed (m/s). In

this case the setpoint is 50 meters and the output are zero meters at the beginning. When the setpoint and feedback signal is differentiated at that point the error signal will be 50. Multiplying the error with the constant Kp it will yield an actuating signal of 5 m/s. As the

distance decreases the error signal will become smaller and thus the actuating signal will too until the robot walks the 50 meters and the error signal is zero.

Steady-state error is a value different from the setpoint that the plant rest on. This happens when a system needs some non-zero value to maintain its state. For example, let say that a drone, controlled and actuated by the propeller speed, has a setpoint of 50 meters. To maintain any constant altitude the drones propellers need to rotate at 100 rounds per minute (RPM). If the proportional gain (𝑘_𝑝) is set to two then the actuating signal will be 50 ∙ 2 = 100𝑅𝑃𝑀 which means that the drone will just hover without increasing its altitude and the steady-state error will be 50 meters. If the proportional gain is increased to five, then the steady-state error will be 20 meters because when the drone increases its altitude and the error decreases so will also the propeller speed decrease to a point where the remaining error multiplied by the proportional gain will equal 100 RPM which is the RPM required for the drone to be able to hover , 20 ∙ 5 = 100𝑅𝑃𝑀.

10 Figure 2.4. Integral part, error vs actuating signal.

At first the drone’s height will oscillate between negative and positive error, these negative and positive values will add and subtract to the previously summed values until the error rests at zero and then the actuating signal will stay constant at 100 RPM to maintain the altitude.

The derivative part of the controller is then used to avoid the oscillations created by the integrator.

Figure 2.4 represents the integral part of a PID controller. Each green and cyan line is multiplied with a pre-determined integral constant (KI) and summed to create the actuating

signal. This is of course happening many times per second.

The derivative part of the controller predicts the future behavior of the error, see figure 2.5. The output that the derivative part output depends on the rate of change of the error with respect to time. That way it can predict the value of the remaining error by calculating how fast the error decreases thus making it more responsive.

11

In figure 2.5, wherever the error signal is constant the derivative actuating line will be horizontal yielding a value of zero. If the error signal is decreasing the slope of the derivative will be negative and thus yielding a negative actuating signal when multiplied with a pre-determined derivative constant (KD) and vise versa. The steeper the slope is the bigger the

value of the derivative.

All the parts of the PID controller have a gain constant which makes the controller tunable to optimize for different plants. The sum of all the parts in the PID controller makes the actuating signal

𝑢(𝑡) = 𝑘_𝑝∙ 𝑒(𝑡) + 𝑘_𝑖 ∙ ∫ 𝑒(𝜏)𝑑𝜏₀𝑡 + 𝑘_𝑑∙𝑑𝑒(𝑡)

𝑑𝑡 (3)

where e(t) is the error as a function of time.

Tuning the PID controller to obtain the desired behavior is probably the biggest part of building such a controller. There have been different tuning methods over the years. The first method in tuning PID controllers was introduced in the 40’s by John G. Ziegler and Nathaniel B. Nichols and it is called the Ziegler-Nichols method. In simple words, the method starts by setting all the gains to zero and start increasing the proportional gain (𝑘𝑝) until a stable oscillation is obtained in the response of the plant with no divergence. Then the P, I and D gains are tuned depending on the oscillation frequency, 𝑇𝑢, and ultimate gain, 𝐾𝑢, from the previous step. The ultimate gain is defined as 1/M where M is the amplitude ratio [14]. What also needs to be considered is the hardware that will be used with the controller and preferably decide on parts before starting with the tuning of the controller. If for some reason the hardware is changed after the tuning is done there is a possibility that the system will behave differently and thus requiring further tuning. [16]

Chapter 3: Methodology

In this chapter the experimental setup will be presented along with the function of the system and how it works.

3.1 Experimental setup

The experimental setup consists of 3different parts, the prototype, the controller and the regulating actuator. For the first part, there is the small-scale prototype which consists of a frame, suspension and a platform for the weights. The frame is proportional to the semi-trailers chassis to try to keep the dynamics as similar as possible. The determined scale was 17% of the original blueprints of the semi-trailer. The suspension consists of 4 syringes with a volume of 50 ml that will replicate the air suspension on the semi-trailer. Those are also proportional to the real air suspension of the semi-trailer at 17% scale. The biggest reason to why syringes is being used is because a big part of the controlling software relies on the pressure from each side of the frame. That way different calculations can be done, for example total mass, lateral weight distribution, total torque on the roll center etc. The syringes similarities are just the scale down of the diameter, otherwise there are no similarities concerning spring rate or dampening. The platform on top of the syringes is used to place the weights and distribute the weight on all four syringes.

12

The total weight that will be applied on top of the platform will be 35 kg at maximum, which translates to around 1.6 bar on each side of the frame if evenly distributed. The calibration of the pressure sensors was necessary to convert the ADC values to bar. The simple equation for pressure was used

𝑝 =𝐹

𝐴 (4)

where F is the force in Newtons and A is the area where the force acts upon. The area was calculated by the diameter of the syringes, approximately 27mm, and the force calculated by the known weight applied on the syringes.

Figure 3.1. BPS130-HG100P-3S pressure sensor.

The next part of the experimental setup is the regulating actuator which is a motion platform from Motion Systems Next Level Racing, see figure 3.2. As the name suggest, this motion platform is intended for simulating in-game motions of racing cars, but the company also offers SDK licenses for developers to build their own software and use the motion platform to apply the custom software. The license also supports MATLAB and Simulink implementation. The prototype is mounted on the motion platform as shown in figure 3.3.

Figure 3.2. Next Level Racing motion platform V2.

13 Figure 3.3. Motion platform with prototype mounted.

This is the first stage to obtaining reasonable values with just the pressure sensors as inputs to the controller, see figure 3.4. The next stage will include accelerometer, gyroscope and probably height sensor in form of potentiometers. The accelerometer and gyroscope data will be fused to obtain the angle relative to the road and use the gyroscope data to also measure angular velocity. The height sensors will also be used to calculate angel but relative to the frame instead of the road.

Figure 3.4. Simple model to regulate the pressures.

14

I2C communication. Both accelerometer and gyroscope are connected to a single breakout board that connects to the Arduino UNO through a breadboard.

Figure 3.5. Arduino UNO dev. board.

Figure 3.6. Wiring for pressure sensors

The control software was developed in MATLAB Simulink and will also be optimized in the same platform. With support of many libraries it was possible to write and read from all hardware that was used in this thesis.

Figure 3.7. Complete Simulink model

15

the model is different subsystems conducting different calculations depending on different constants and different signals. In the first experiment, it was not necessary to use all the input signals and subsystems just to eliminate as many variables as possible and make it easier to troubleshoot eventual problems that would occur.

The architecture of the software was pre-determined in a form of a simple flow chart. How the different input signals were treated was discussed parallel to building the software. The library and toolbox for Simulink to support Arduino UNO hardware was free to install from the add-ons tab in MATLAB. The library and toolbox for the motion platform was downloaded and installed to Simulink with a purchased license from the company that provided the motion platform, Motion Systems.

To be able to run a real time simulation with hardware, some changes were required in Simulink. To do those changes the “Model setting” was accessed through the “Modeling” tab in Simulink. From there the “Hardware Implementation” tab was accessed in the pop-up window, see figure 3.8.

Figure 3.8. Configuration window for hardware implementation

If the Arduino hardware support package for Simulink is installed in MATLAB, then there should be an option in the “Hardware board” for an Arduino UNO board. After that the tab “Target hardware resources” was opened and after that the “Host-board connection” tab as well. In there, the “Set host COM port” was selected to “Manually” and the COM port number to the Arduino was written in the writing tab bellow.

To check for the COM port of the Arduino the control panel for Windows was opened and searched for “device manager” in the search tab. In device manager the “Ports (COM & LPT) tab was accessed to find the Arduino COM port. This enabled real time simulation with hardware implementation.

16

order to enable the platform, there must be a constant value of one to the “enable” input of the block. The servos that control the pitch and roll movements are of 16-bit resolution. The decimal value of 16-bit is 65,536. That means that it takes 65,536 steps to complete a full motion from one side to the other. The input range for this motion platform is between -32,767 to -32,767. An input of zero will keep the motion platform straight, a positive value will move it to one side and a negative value to the other with the maximum values -32,767 and 32,767.

To keep the input between these two values the actuating signal needs to be between -1 and 1 to later gain the signal with 32,766 before sending it to the motion platform input block. To accomplish that, the full weight of 35kg was placed on two syringes on one side of the prototype to get the maximum output from the pressure sensor with the maximum weight. The maximum output value is then used to divide the difference between the two sensors to obtain values between -1 and 1. A saturation block was implemented as a last precaution to restrict the actuating signal between -1 and 1.

Figure 3.9. Input block for motion platform.

3.2 Block diagram

Figure 3.10displays the block diagram for the overall function of the system. It starts with the semi-trailer which outputs four different signals with the help of sensors. Those signals are 3-axis acceleration, 3-axis angular rate and pressure in the air suspension for each side. The accelerometer and gyroscope data get processed and fused to get the angle relative to sea level. The pressure signals get processed to output the values in bar.

17

As mentioned before the pressure sensors are outputting analog signals between 0,5-4,5V and the Arduino has a 10-bit ADC, 1024 steps, from 0-5V. To obtain the right pressure values in bar the following calculations were made.

0,5

5 ∙ 1024 ≈ 102 (5) 4,5

5 ∙ 1024 ≈ 922 (6)

At zero pressure the ADC value will be at 102 and at max pressure for the pressure sensor the ADC value will be 922 which corresponds to 6,89 bar.

6,89

(922−102)≈ 0,00840 𝑏𝑎𝑟/𝑠𝑡𝑒𝑝 (7)

Now that the coefficient that will be multiplied with the ADC value is known the only thing remaining to do is offset the ADC value so that it is zero at zero pressure and multiply with the coefficient obtained from equation (4) and convert bar to pascal to be able to, later on, calculate forces.

(𝐴𝐷𝐶 − 102) ∙ 0,00840 ∙ 100 000 = 𝑝 [𝑃𝑎] (8)

Figure 3.11. Semi – trailer diagram.

Figure 3.11 displays a simple sketch of a semi-trailer with some determined to know values. Some of the values are known. Those that are calculated in the system are θ, mg, α, moment arm for mg, ln, and hCG, which is also the moment arm for the acting lateral acceleration.

The vertical lines B and C are the theoretical limits where if CG passes beyond, then roll-over is inevitable. Point A is where the semi-trailer will roll about assumed that it is a stiff body. To calculate the weight distribution in lateral translation the following calculations were made.

𝑙_𝐿 = 𝑝𝐿

18

𝑙𝑅 = 𝑝𝑅

𝑝𝐿+𝑝𝑅∙ (𝑙𝑤𝑖𝑑𝑡ℎ) (10)

Equations (6) and (7) calculate the distance of the center of gravity (CG) from each side. 𝑙𝑤𝑖𝑑𝑡ℎ is the distance between the mid-point from each wheel.

To calculate the total mass of the load equation (1) was used and solved for F. The area was calculated by measuring the diameter of the syringe. Because the weight is distributed across four syringes the area got multiplied by four.

𝐹 = 𝐴 ∙ 𝑝 (11) 𝐴𝑡𝑜𝑡 = 4(𝑟2∙ 𝜋) (12) 𝑚 =𝐹 𝑔 (13) (8) and (9) in (10) yields, 𝑚 =4(𝑟2∙𝜋)∙𝑝 𝑔 (14)

To be able to calculate the Acceleration torque and angle threshold, the height of CG needs to be known. In this situation there is not a simple way to obtain that. That’s why the height of CG was calculated proportionally to the mass. The torque that is created by the lateral acceleration:

𝜏_𝑎𝑐𝑐 = ℎ_𝐶𝐺∙ 𝛼_𝑙𝑎𝑡 (15) The torque that is created by CG:

𝜏_𝐶𝐺 = 𝑙_𝑛∙ 𝐶𝐺 (16)

𝑙_𝑛 depends on what direction the acceleration acts. If the acceleration acts as shown in figure 3.11then CG is multiplied by 𝑙_𝑅 because the roll point is on the right side. The angle threshold θ is calculated for both sides, the equation goes as follows:

𝜃𝑛 = tan−1( 𝑙𝑛

ℎ𝐶𝐺) (17)

As the CG shifts from right to left and vice versa the threshold angels will change thus needing to be calculated again. To do that, a zero lateral acceleration is needed to be certain that the cargo is not shifting position when calculating these angles. These angles will be compared to the actual angle of the semi-trailer chassis and controlled by a PID controller to a given setpoint.

The CG torque and acceleration torque will also be differentiated and controlled by a PID controller to a given setpoint.

19

Chapter 4: Result

In this chapter the results of the experiment will be presented with figures and graphs showing the response of the PID controller with high intensity and low intensity disturbances. One could relate that kind of disturbance as avoiding maneuvers when driving or simply turning in a curve with excessive velocity. The experiment showed that with relative normal gains in any of the two PID controllers the motion platform became unstable and oscillated uncontrollably. The low PID controller’s gains was KP = 0.2, KD = 0.005 and KI = 0 and for the

high PID controller the gains was KP = 1, KD = 1 and KI = 0. The results for the low intensity

disturbances seen in figure 4.2 below display a fast response with low transient oscillation in the system and in the high intensity disturbances seen in figure 4.1 the system is also fast in response and stable with few oscillations when trying to equalize the pressures.

Figure 4.1. Diagram for high gain PID response for high disturbance. Highest peak is the moment the disturbance is applied.

20

Figure 4.2. Diagram for low gain PID response for low disturbance. Highest peak is the moment the disturbance is applied.

In figure 4.3 the angle of the motion platform is shown and clearly seen that for the lower intensity disturbances there are less counter action versus the high intensity disturbances which yield higher roll movement of the motion platform to counter act the difference of pressure in the syringes.

21

Chapter 5: Conclusion & future work

This thesis has provided an experimental evaluation of the roll over prevention system that has the ability to counter act the roll tendencies of the prototype when outer disturbances are applied, which corelate to lateral accelerations with result of torque on the roll center of a semi-trailer. The obtained results from the experimental setup are satisfactory when it comes to this particular experiment proving that the developed controller can handle both low and high amplitude disturbances with low transients, which proves that the system is fast but stable as well. The control signal did not utilize the full range of the motion platform and that is because with higher gains in the PD controller the system became unstable and started oscillating uncontrollably. That can depend on the motion platforms step response, which means that it takes some time for the platform to adjust to the new values given by the PD controller and that lag in time created the oscillations. The platform was quick to respond and settle to low degrees of roll but not so much for higher degrees.

The experimental setup revolved around a stationary experiment with one set of pressure sensors and a motion platform as the actuator capable of roll rotation to a certain degree. Future work and studies can utilize a combination of the pressure sensor used in this thesis and other type of sensors and experiment in a non-stationary scenario where other dynamical factors can influence the dynamics of the system.

More future work can include developing a brake system for the semi-trailer that will allow both the ABS module and the roll over prevention system to actuate the brakes with the ABS module being the default.

22

References

[1] https://en.wikipedia.org/wiki/California_Air_Resources_Board [2] https://en.wikipedia.org/wiki/United_States_Environmental_Protection_Agency [3] https://en.wikipedia.org/wiki/Automotive_safety#History [4] https://en.wikipedia.org/wiki/Vehicle_emissions_control#History [5] https://en.wikipedia.org/wiki/United_States_Department_of_Transportation [6] https://www.allpar.com/cars/desoto/electrojector.html [7] https://www.freno.se/ [8]https://www.wabco-auto.com/americas_en/Our-Solutions/Trailer-solutions/Trailer- Safety/Trailer-Control-Systems/Trailer-Control-Systems [9] https://www.gwtrans.com/the-history-of-semi-trailer-trucks/ [10] https://www.acea.be/uploads/publications/factsheet_trucks.pdf [11] https://www.daf.com/en/sites-landing [12] https://www.iveco.com/sweden/pages/homepage.aspx [13] https://kamaz.ru/en/

[14] Optimum Settings for Automatic Controllers by J.G. ZIEGLER and N. B. NICHOLS