Automotive Engineering MEng Projects

An autonomous system is tasked with localization, mapping, path planning and obstacle avoidance. A lot of research and efforts have been made to achieving autonomy of autonomous boats. RobotX challenge is a competition created to challenge and educate students to develop maritime autonomy through a systems engineering approach. Students are challenged to develop innovative solutions to ensure that the autonomous boat performs certain tasks in any environment. The entire project is done in a virtual setting using the Robot Operating System. This project deals with the mapping and navigation of the autonomous boat in a pre-defined environment.

The Robot Operating System provides collaborative and open source solutions that promote the development of robust and complete products. The project uses various developed packages and creates an interface to enable the autonomous navigation of the boat. Using the available lidars, an octomap is created to map the world. This octomap is then used to perform the guided navigation. The project uses the navigation stack of the Operating System. Then, the Dijkstra’s algorithm is used as the global planner to find a path from the start point to the goal point. This algorithm uses the 2-D map to identify the shortest path from the start point to the goal point. A local planner is also implemented. This planner uses the available 2-D map along with the information from sensors to identify obstacles and help the autonomous boat find a path that is along the global path but help avoid any dynamic obstacles. The project focuses on using the existing packages and tuning them to create a robust solution to the tasks in hand.

As little attention has been paid to the vehicle passive safety systems in the automated vehicle industry, this research aims to study the occupant restraint system performance in a new automated vehicle model designed by Nimbus. We utilized LS-DYNA and HyperMesh to simulate full frontal and frontal oblique impacts of the vehicle with advanced seat belt and airbags. We also improved the restraint systems to reduce injury measures in the crashes. We found that in a full frontal impact, a softer airbag could reduce the HIC15 value to 82% and chest acceleration to 83% compared with the baseline, while a collapsible steering column reduced HIC15 to 71% and chest acceleration to 69%. In the 30° oblique impact, presence of the curtain airbag reduced HIC15 to 56% and chest acceleration to 70% compared with the model not equipped with a curtain airbag. We may conclude that a softer driver airbag and a collapsible steering column are effective in mitigating serious head and chest injuries in the full frontal impact. Curtain airbag can greatly reduce the head and chest injury risk when the occupant has the potential to roll off the driver airbag. Therefore, we recommend installation of a smart airbag that can control the airbag vent size based on the occupant weight, a collapsible steering column, and curtain airbags in the automated vehicle model to greatly reduce the occupant injury risks in the crashes.

The project consisted of creating test cases for testing features of Infotainment system and then stress testing the system for stability. Test cases were created for features like Lane assist, blind spot monitoring, navigation etc., The testing took place in steps, starting with unit testing where individual modules were tested for there. functionality before integrating them and testing the whole system. One of the modules that was tested by me was the performance of machine learning algorithm YOLOv3 used for object detection. The performance was enhanced from 35 mAP to 50mAP by using Non maximum Suppression technique.

Other than that, the whole system was tested was tested for a total of 2500 hours, executing scripts 24/7 for three months. A total of 360 features has been tested for stability and 20 crashes has been discovered and fixed. The testing phase is still underway but on different vehicle built, the features from old vehicle system testing were transferred to a different one with little tweaking to the whole programme.

Electric vehicles are becoming ubiquitous by the day. An important parameter for measurement is the battery’s State of Charge to get an optimal performance from the electric vehicle. It is a measure of the average lithium concentration in the electrode. In layman’s language it can be understood as analogous to fuel. However, unlike conventional fuel tanks, one cannot directly measure the SOC and therefore, is estimated. In this project, a battery was characterized and parameterized using double RC equivalent circuit model. Kalman filtering techniques such as EKF and UKF were used to estimate the SOC. Particle filter algorithm was developed and used to estimate the SOC and performance of the three filters were compared based on accuracy and computation expense.

In hopes of becoming an industry leader in additive manufacturing, GM built a brand-new 3D printing lab called the Additive Industrialization Center (AIC) designated for 3D printing tools and parts. Our project team was tasked with capturing the cost to manufacture production parts within the AIC and comparing those costs to outsourcing 3D printed parts to a supplier. Our project goal was to show that the cost to internally manufacture would show a net benefit to the total enterprise. Our project focused on a new low-volume vehicle program to use as a template for all future additive production.

Our team has worked to deliver course material for a new foundational Vehicle Dynamics class starting the Winter 2021 semester at the University of Michigan. The overall objective of this project is to provide the course curriculum and technical foundation for this upcoming Vehicle Dynamics class by integrating coursework and lecture material with interactive vehicle-level modeling using industry-standard tools such as SIMULINK, CarSim, and Cameo Magic Draw to develop functional models. The emphasis is on developing the curriculum, assignments, and models in an interactive and engaging format that allows students to visualize the physical vehicle interactions in order to gain an understanding of Vehicle Dynamics concepts.

Over the course of this semester, our team has developed learning objectives as well as both CarSim and SysML models for each of the course’s ‘Sectors’ (Accelerate, Brake, Ride, Corner, etc.) to further reinforce key concepts to the students who take this course. The largest value-add benefits are seen in the University of Michigan adding a new course to its graduate studies curriculum and reinforcing its presence in the world of Automotive Engineering education, while also supplying students with an incredible opportunity to prepare themselves for employment in the Automotive field through use of the knowledge, tools and skills used in industry.

This project examined hardware sensing solutions intended to detect the conditions inside an electric vehicle battery pack during a single cell thermal runaway condition. Various sensors were proposed and compared, with the highest performing sensor selected and recommended for use in the production intent design. Over the course of the project, eight total sensors were evaluated analytically and tested with physical hardware, with the highest performing sensors recommended to the GM product development team. The expected benefit of this project is to reduce development time compared to alternative production intent solutions by up to 18 months, with a 45% to 50% cost savings.

One source of energy loss in automatic transmissions is a disengaged friction plate clutch that is not used in certain gears. The current production design has several elements in order to minimize the amount of energy lost to the clutch. However, one of the components is 28% of the total cost of the clutch assembly. In order to reduce the cost of the clutch assembly without decreasing the efficiency of the transmission, a Design for Six Sigma project was launched to systematically investigate the effect each design factor has on minimizing energy loss in the clutch. The goal is to find a set of design factors that maintain or reduce the baseline’s amount of energy loss at a minimal cost.

The COVID-19 pandemic prevented completion of all the tests by the time of the report submission. However, the baseline design was tested at several combinations of different automatic transmission fluid temperatures, clutch clearances, and automatic transmission fluid flow rates. It was found that the amount of energy loss for the baseline design could vary as much as 36% from the best to worst set of conditions. The flow rate was the most significant factor to increasing or decreasing energy loss to the clutch and could vary the performance 20% on its own. While the fluid temperature and clutch clearance only varied performance by 11% and 8%, respectively. The importance of this is that we now know a specific parameter to investigate further to see if a change can be made that will consistently reduce the amount of energy loss to the clutch. Once the project resumes, it is likely that an optimized design can be found that minimizes the energy loss for a minimal cost.

One of the challenges as the auto industry moves to all electric vehicles is the battery pack weight increase. Battery packs in electric vehicles tend to increase the weight of the vehicle by 500 kg. As a result of this increase, the torsional stiffness of an electric vehicle needs to be 150% off an internal combustion vehicle to maintain proper noise, vibration, and ride harshness properties. The goal of this project is to evaluate various vehicle body structure variables that can contribute to an increase in torsional rigidity including but not limited to, increasing the number of mounting points, sheet metal thickness, and rocker beam thickness in an electric vehicle body structure. This project demonstrated that above 16 bolt connections between the vehicle cabin and battery pack there was no benefits in terms of cost, as stiffness increase was minimal and extra connections would only add to the manufacturing costs in the assembly plant. The best variable that increased the overall torsional stiffness of the vehicle was to increase the vehicle cabin sheet metal thickness by 0.2 mm at the smallest total material cost of $17.87 per vehicle. Varying the rocker beams for both the cabin and battery box were the worst in terms of cost, price increase of $245 for battery pack rocker. The rocker beams should only be sized to meet vehicle safety requirements for side impact and battery pack protection from punctures. The benefit in the findings are that the vehicle cabin has been well researched over the life of the auto industry. Current vehicle cabin stiffening methods can then be similarly be used to increase the torsional stiffness of an electric vehicle.

Currently, each engineer uses different methodologies to analyze and interpret warranty data, and to conduct root cause analysis. There are also different programs and formats used to present the data to others, which is inconsistent and may be confusing. Identifying the root cause with confidence more quickly allows the solution to be released into production more quickly, reducing the financial impact to the company and improving customer satisfaction.

Best practices and improvement opportunities were gathered from reviewing previous projects and from interviewing managers within the organization. The introduction of standard processes and templates for conducting warranty analysis and root cause analysis reduces the time that engineers spend in completing each activity. The estimated total reduction in time to conduct warranty analysis and root cause analysis activities is approximately a 24% reduction over the current state.

Truck customers are looking for the capability to tow or haul their toys, campers, boats, horses, RVs, and much more. Capability is key for truck customers, but with that added capability comes a matching need for safety, particularly an avoidance of overloading their vehicle. I set out to find a method for estimating a truck’s Gross Combined Weight (GCW – combination of truck, payload, and/or trailer weight) using data readily available via Ford’s OpenXC open-source platform, and developing an Android app to communicate via Bluetooth with the vehicle. The app, MyFordWeights, reads in powertrain data and runs an algorithm to estimate GCW based upon the transmission torque and vehicle acceleration. As a proof-of-concept, the goal was to achieve an estimated weight accuracy within 250lbs of the measured GCW.

On average, the app was able to deliver to that accuracy range, however there were still instances of measurements as inaccurate as up to 700lbs in some weight ranges. To address this, the algorithm could be altered with a more effective means of tracking vehicle acceleration to provide a more reliable form of estimating GCW. In the testing phase, it was discovered that the vehicle acceleration estimation in the algorithm was the most inconsistent of the data points. Once the algorithm is to a point of 100% consistency at all weight ranges, the Android app could provide a real benefit to pickup truck owners who are looking for more confidence in their towing and payload-carrying abilities.

With increasing fuel prices and stringent emission standards on the automotive industry, manufacturers are seeking to improve their fuel economy and overall vehicle efficiency in order to keep pace with the evolving landscape. For this project three waste heat devices were analyzed to see how they could help improve fuel efficiency. Ultimately, the Rankine cycle was selected as the waste heat recovery device to implement on a model based vehicle simulation. The addition of a Rankine cycle on a baseline naturally aspirated engine led to an approximate 2% improvement in fuel economy on the Federal Test Procedure 75 drive cycle.

Automotive Radar Sensors are an integral part of the Advanced Driver Assist Systems. Development of algorithms to process radar signals is important to understand and characterize the environmental conditions and obstacles. Development of several hardware to test new algorithms or designs can be very expensive. The systems can be modelled and simulated to test new designs and algorithms, which will save a lot of time and cost involved in the development. This will in turn reduce the cost of the sensor as well as accelerate the development of the systems.

By addressing the identified drawbacks of the reference model, the proposed model has more flexibility to simulate a variety of scenarios for automated driving and object tracking which were out of the scope of the reference model.

This project was aimed to complete the groundwork required to convert a Ricardo Hydra engine to a Diesel Engine configuration so that it could be used for research on advanced combustion techniques. In this project I was able to model the Hydra Engine and a Jug that can be used to adapt to the cylinder head of the donor GM Diesel Engine. We plan to machine the jug at the lab and complete the conversion over the coming weeks to begin advanced combustion studies.

Isuzu Motors Ltd., a global leader in commercial vehicles and diesel engines approached the University of Michigan’s MDP team with a problem that they noticed with their transmission synchronizer rings. The transmission system is a vital part of an automobile’s powertrain and maximizes the power of the vehicle by allowing higher speed at lower engine revolutions while providing the torque required for acceleration. In order to ensure smooth and effective operation of manual transmission systems, synchronizers are used to equalize the speed of gear wheel and transmission shaft during the change of gears by using friction. Failure of synchroniser rings will often result in loss of friction and generate a lot of noise and induces extensive wear of the gears due to grinding. High operating temperatures may also lead to the deterioration the materials in the ring which would eventually result in failure of the ring. It was noticed by Isuzu’s testing and quality engineers that the synchronizer rings in one of their transmission systems failed prematurely. An internal study was conducted, and it was hypothesized that degradation of one of the materials was leading to this failure. The purpose of this project was to characterize this failure from a materials point of view and find out its root-cause.

Start Stop is a fuel saving technology targeted at reducing unnecessary engine idle time by shutting down the engine and automatically re-starting it when requested by the driver or vehicle. This technology requires that the engine software be calibrated to deliver smooth engine restarts with minimal disturbance to the driver and vehicle occupant. Customer satisfaction with the technology has been an area of concern, in part due to engine restart vibrations perceived by the customer. This disturbance can be minimized though optimization of the powertrain calibration with a goal that is twofold: first, the engine must start as quickly as possible to provide power requested by the driver, and secondly, vibrations felt by the vehicle occupants must be minimized. In order to achieve these goals, a structured design of experiments methodology was used to test multiple factors simultaneously while observing key engineering metrics for restart smoothness and speed. A screening experiment and a spark-focused experiment conducted in this project found recommended improvements for controlling spark and throttle, but primarily found spark timing to cause the most variation in engine restart smoothness and speed. The results of these experiments propose a calibration method that focuses on using design of experiments to optimize the calibration in an ordered approach. These results that deliver smoother restarts on a more consistent basis hope to translate to improved customer satisfaction and acceptance of Start Stop technology.

Ford Motor Company views this project as an opportunity to investigate ways to increase steering angle controller performance. An increase in performance would allow the autonomous vehicle platform to better hold to angle requests and a path sent by a virtual source. It is also useful to investigate the computational requirements of these methods; higher computational costs would likely cause MCU runtime faults, which is not recommended for such a safety critical function. A high computational cost would also drive Ford to increase micro controller capabilities, which translates, to higher costs. This project found that there are methods in which to increase angle performance by utilizing intermediate function calls between angle requests, some methods being more advantageous. This project also investigated the computational efficiency of these methods.

The purpose of the project is to improve the traditional sideslip estimation and correction based ESC algorithm found in light passenger cars using additional information available to an autonomous vehicle. Ideally, an AV should be able to outperform a traditional driver, similar to how a vehicle with ESC can outperform a vehicle with only four wheel ABS braking. The results of this project show a proof of concept ESC algorithm that uses additional inputs for control of individual wheel pressures to reduce error in target heading angle/yaw rate error, which keeps the vehicle on trajectory. The value added benefit is being able to react to scenarios such as an object darting out in front of the vehicle, or split mu braking where asymmetric braking can help rotate and steer the vehicle around an object or back to the intended path.

Upon receiving the first prototype vehicles for a new vehicle program at a major automotive manufacturer, it was found that the new transmission has a hardware issue that causes a drive ability issue. The root cause was found to be a torque converter that was not adequately stiffened for the transmission’s hydraulic circuit. In order to mitigate this issue, the engineering team investigated all possible solutions. A hardware change was only moderately successful at reducing the drivability issue, and it would cost approximately $2.075 million (in the first year of production alone) to implement, the delayed timing would not meet the program timeline, and this new design added a significant amount of weight. In order to avoid this high cost hardware change, the team came up with the idea to create a compensation software to combat the torque converter compliance issue. The engineering team was able to create/validate software in order to successfully mitigate the torque converter issue and for much less cost. After checking on multiple calibration vehicles, the success rate of reducing this flare to an acceptable level is approximately 99.4% which was seen as a success by the engineering management.

Due to increasing concern over the societal cost of CO2 emissions, countries around the world have been pushing toward increasingly more stringent fuel economy and emissions regulations for the automotive sector. This added stringency has led automakers to deploy significant resources to develop electrified powertrains to meet the regulatory requirements. This project was developed to assist Fiat Chrysler Automobiles LLC in studying optimization opportunities on a selected electrified powertrain configuration. Optimization opportunities were identified by sweeping different motor sizes, final drive ratios, and operating system voltages to identify the configuration that provided the most improvement over the baseline system. As a result, twelve different configurations were evaluated against the selected metrics with a leading candidate (smallest motor with boost capability and base axle ratio) selected through the use of a weighted decision matrix. This leading candidate was then compared against additional alternate architectures under consideration to further evaluate its benefits and tradeoffs. Ultimately, the leading optimization candidate and the best performing alternate architecture were thoroughly studied against the baseline with the leading optimization candidate providing a 5.0% improvement in unadjusted combined fuel economy and a 5.5% improvement in CO2 while the best performing alternate architecture provided a 2.3% improvement in fuel economy and a 4.6% improvement in CO2. It is recommended that both systems are investigated further to confirm the observed benefit as well as evaluate any additional opportunities for optimization

I currently lead a small powertrain engineering team at Akka Technologies, working on Maserati/Alfa Romeo projects. This team is brand new and consists of only myself and a fresh college graduate. Because of these factors, I undertook this project to both generate relevant technical documentation as well as level up my coworker’s abilities so that we can work more efficiently as a team.

At the conclusion of this project, our team was able to improve manpower utilization efficiency by 36%, increase our workload by 33%, and potentially improve on-boarding efficiency by 50%. The increased productivity that resulted from this work will result in an estimated first year financial benefit of $132,000, a significant personnel budget savings.

The objective of this project is to develop a vision based SLAM approach and evaluate its effectiveness of performing low speed autonomous navigation of unknown parking lots. For this project, we researched, formulated, and implemented a SLAM algorithm in C++ with production intent software constraints in mind. We tested the algorithm in simulation, on a publically available dataset and found that incorporating vision based techniques into the SLAM method is beneficial. Our main value-add is developing a new algorithm for Ford.