Search Results

This paper documents the processes and methods used by the Ford GT team to meet aerodynamic targets. Methods included Computational Fluid Dynamics (CFD) analysis, wind tunnel experiments (both full-size and scale model), and on-road experiments and measurements. The goal of the team was to enhance both the high-speed stability and track performance of the GT. As a result of the development process, significant front and rear downforce was achieved while meeting the overall drag target.

The Ford GT Program Team was allocated just 22 months from concept to production to complete the Electrical and Electronics systems of the Ford GT. This reduced vehicle program timing - unlike any other in Ford's history -- demanded that the team streamline the standard development process, which is typically 54 months. This aggressive schedule allowed only 12 weeks to design the entire electrical and electronic system architecture, route the wire harnesses, package the components, and manufacture and/or procure all components necessary for the first three-vehicle prototype build.

The Ford GT powertrain (see Figure 1) is an integrated system developed to preserve the heritage of the LeMans winning car of the past. A team of co-located engineers set out to establish a system that could achieve this result for today's supercar. Multiple variations of engines, transaxles, cooling systems, component locations and innovations were analyzed to meet the project objectives. This paper covers the results and achievements of that team.

The High Voltage Power Distribution Unit (PDU) is constructed of magnesium in support of Fuel Cell Electric Vehicle (FCEV) weight reduction efforts. The PDU distributes and controls a nominal 75 kilowatts of power generated by the Fuel Cell, the primary source of High Voltage power, to all the vehicle loads and accessories. The constraints imposed on the design of the PDU resulted in a component highly susceptible to general and galvanic corrosion. Corrosion abatement was the focus of the PDU redesign. This paper describes the redesign efforts undertaken by Ford personnel to improve the part robustness and corrosion resistance.

The proposed technique is a tailored deep neural network (DNN) training approach which uses an iterative process to support the learning of DNNs by targeting their specific misclassification and missed detections. The process begins with a DNN that is trained on freely available annotated image data, which we will refer to as the Base model, where a subset of the categories for the classifier are related to the automotive theater. A small set of video capture files taken from drives with test vehicles are selected, (based on the diversity of scenes, frequency of vehicles, incidental lighting, etc.), and the Base model is used to detect/classify images within the video files. A software application developed specifically for this work then allows for the capture of frames from the video set where the DNN has made misclassifications. The corresponding annotation files for these images are subsequently corrected to eliminate mislabels.

A study was conducted using Ford's nine standard CFD calibration models as described in SAE paper 940323. The models are identical from the B-pillar forward but have different back end configurations. These models were created for the purpose of evaluating the effect of back end geometry variations on aerodynamic lift and drag. Detailed experimental data is available for each model in the form of surface pressure data, surface flow visualization, and wake flow field measurements in addition to aerodynamic lift and drag values. This data is extremely useful in analyzing the accuracy of the numerical simulations. The objective of this study was to determine the capability of a digital physics based commercial CFD code, PowerFLOW ® to accurately simulate the physics of the flow field around the car-like benchmark shapes.

Ford Motor Company has recently implemented a Hardware-In-the-Loop (HIL) testing system for a new, highly complex, hybrid electric vehicle (HEV) Electronic Control Unit (ECU). The implementation of this HIL system has been quick and effective, since it is based on proven Commercial-Off-The-Shelf (COTS) automation tools for real-time that allow for a very flexible and intuitive design process. An overview of the HIL system implementation process and the derived development benefits will be shown in this paper. The initial concept for the use of this HIL system was a complete closed-loop vehicle simulation environment for Vehicle System Controller testing, but the paper will show that this concept has evolved to allow for the use of the HIL system for many facets of the design process.

In future vehicles (e.g. fuel cell vehicles, hybrid electric vehicles), the electrical system will have an important impact on the mechanical systems in the car (e.g. powertrain, steering). Furthermore, this coupling will become increasingly important over time. In order to develop effective designs and appropriate control systems for these systems, it is important that the plant models capture the detailed physical behavior in the system. This paper will describe models of two electrical components, a battery and a supercapacitor, which have been modeled in two ways: (i) modeling the plant and controller using block diagrams in Simulink and (ii) modeling the plant and controller in Dymola followed by compiling this model to an S-function for simulation in Simulink. Both the battery and supercapacitor model are based on impedance spectroscopy measurements and can be used for highly dynamic simulations.

Combustion in modern spark-ignition (SI) engines is increasingly knock-limited with the wide adoption of downsizing and turbocharging technologies. Fuel autoignition conditions are different in these engines compared to the standard Research Octane Number (RON) and Motor Octane Numbers (MON) tests. The Octane Index, OI = RON - K(RON-MON), has been proposed as a means to characterize the actual fuel anti-knock performance in modern engines. The K-factor, by definition equal to 0 and 1 for the RON and MON tests respectively, is intended to characterize the deviation of modern engine operation from these standard octane tests. Accurate knowledge of K is of central importance to the OI model; however, a single method for determining K has not been well accepted in the literature.

As part of the International Lubricant Standardization and Approval Committee's (ILSAC) goal of developing a global automatic transmission fluid (ATF) specification, members have been evaluating test methods that are currently used by various automotive manufacturers for qualifying ATF for use in their respective transmissions. This report deals with comparing test methods used for determining torque capacity in friction systems (shifting clutches). Three test methods were compared, the Plate Friction Test from the General Motors DEXRON®-III Specification, the Friction Durability Test from the Ford MERCON® Specification, and the Japanese Automotive Manufacturers Association Friction Test - JASO Method 348-95. Eight different fluids were evaluated. Friction parameters used in the comparison were breakaway friction, dynamic friction torque at midpoint and the end of engagement, and the ratio of end torque to midpoint torque.

Pressure Sensitive Paint (PSP) has been used for several years by the aircraft industry in transonic wind tunnel testing where the oxygen concentrations are low and the luminescence of the paint is easily recorded. Extending PSP to slower speeds where the oxygen concentrations are closer to atmospheric conditions is much more challenging. For the past few years, work has been underway at both Wright Patterson Air Force Base and Ford Motor Company to advance PSP techniques for testing at slower speeds. The CRADA (Cooperative Research and Development Agreement) provided a way for comparisons to be made of the different PSP systems that were being investigated. This paper will report on PSP tests conducted as part of the CRADA.

As part of an ongoing technical collaboration between Ford and Rouge Steel Company, a comprehensive study of door slam event was undertaken. The experimental phase of the project involved measurements of accelerations at eight locations on the outer panel and strains on six locations of the inner panel. Although slam tests were conducted with window up and window down, results of only one test is presented in this paper. The CAE phase of the project involved the development of suitable “math” model of the door assembly and analysis methodology to capture the dynamics of the event. The predictability of the CAE method is examined through detailed comparison of accelerations and strains. While excellent agreement between CAE and test results of accelerations on the outer panel is obtained, the analysis predicts higher strains on the inner panel than the test. In addition, the tendency of outer panel to elastically buckle is examined.

A correlation of aerodynamic wind tunnels was initiated between Chrysler, Ford and General Motors under the umbrella of the United States Council for Automotive Research (USCAR). The wind tunnels used in this correlation were the open jet tunnel at Chrysler's Aero Acoustic Wind Tunnel (AAWT), the open jet tunnel at the Jacobs Drivability Test Facility (DTF) that Ford uses, and the closed jet tunnel at General Motors Aerodynamics Laboratory (GMAL). Initially, existing non-competitive aerodynamic data was compared to determine the feasibility of facility correlation. Once feasibility was established, a series of standardized tests with six vehicles were conducted at the three wind tunnels. The size and body styles of the six vehicles were selected to cover the spectrum of production vehicles produced by the three companies. All vehicles were tested at EPA loading conditions. Despite the significant differences between the three facilities, the correlation results were very good.

Ride Hailing service and Dynamic Shuttle are two key smart mobility practices, which provide on-demand door-to-door ride-sharing service to customers through smart phone apps. On the other hand, some big companies spend millions of dollars annually in third party vendors to offer shuttle services to pick up and drop off employees at fixed locations and provide them daily commutes for employees to and from work. Efficient fixed routing algorithms and analytics are the key ingredients for operating efficiency behind these services. They can significantly reduce operating costs by shortening bus routes and reducing bus numbers, while maintaining the same quality of service. This study developed an off-line optimization routing method for employee shuttle services including regular work shifts and demand based shifts (e.g. overtime shifts) in some regions.

Self-piercing rivet (SPR) joints are a key joining technology for lightweight materials, and they have been widely used in automobile manufacturing. Manual visual crack inspection of SPR joints could be time-consuming and relies on high-level training for engineers to distinguish features subjectively. This paper presents a novel machine learning-based crack detection method for SPR joint button images. Firstly, sub-images are cropped from the button images and preprocessed into three categories (i.e., cracks, edges and smooth regions) as training samples. Then, the Artificial Neural Network (ANN) is chosen as the classification algorithm for sub-images. In the training of ANN, three pattern descriptors are proposed as feature extractors of sub-images, and compared with validation samples. Lastly, a search algorithm is developed to extend the application of the learned model from sub-images into the original button images.

In order to maximize customer satisfaction in today's global market place, the quality of products and services need to be improved continually. Increased focus on quality, with the attendant proliferation of methods and tools, has created the need for a comprehensive framework to guide the selection of the tools. Individuals within an organization need to know what tools are appropriate in a given situation, and when, where and how the knowledge gained from an effort should be documented. In addition, a common nomenclature to convey quality related information to each other would avoid confusion and improve the communication process thus improving the effectiveness and productivity of the organization. This paper integrates tools that have evolved recently with the old tools that have been in use for a number of years.

This paper presents a design process with data mining technique and advanced optimization strategy. The proposed design method provides insights in three aspects. First, data mining technique is employed for analysis to identify key factors of design variables. Second, relationship between multiple types of size and shape design variables and performance responses can be analyzed. Last but not least, design preference can be initialized based on data analysis to provide priori guidance for the starting design points of optimization algorithm. An exhaust system design problem which largely contributes to the improvement of vehicular Noise, Vibration and Harshness (NVH) performance is employed for the illustration of the process. Two types of design parameters, structural variable (gauge of component) and layout variable (hanger location), are considered in the studied case.

The wide applications of automatic sensing devices and data acquisition systems in automotive manufacturing have resulted in a data-rich environment, which demands new data mining methodologies for effective data fusion and information integration to support decision making. This paper presents a new methodology for developing a diagnostic system using manufacturing system data for high-value assets in automotive manufacturing. The proposed method extends the basic attributes control charts with the following key elements: optimal feature subset selection considering multiple features and correlation structure, balancing the type I and type II errors in decision making, on-line process monitoring using adaptive modeling with control charts, and diagnostic performance assessment using shift and trend detection. The performance of the developed diagnostic system can be continuously improved as the knowledge of machine faults is automatically accumulated during production.

An eight-group taxonomy was created to classify real-world frontal crashes from the Crashworthiness Data System (CDS) component of the National Automotive Sampling System (NASS). Three steps were taken to develop the taxonomy: (1) frontal-impact towaway crashes were identified by examining 1985-2005 model year light passenger vehicles with Collision Deformation Classification (CDC) data from the 1995-2005 calendar years of NASS; (2) case reviews, engineering judgments, and categorization assessments were conducted on these data to produce the eight-group taxonomy; and (3) two subsets of the NASS dataset were analyzed to assess the consistency of the resulting taxonomic-group frequencies. “Full-engagement” and “Offset” crashes were the most frequent crash types, each contributing approximately 33% to the total. The group identified as “D, Y, Z No-Rail” was the most over-represented crash type for vehicles with at least one seriously-injured occupant.

Variant modeling techniques have been developed to allow systems engineers to model multiple similar variants in a product line as a single variant model. In this paper, we expand on this past work to explore the extent to which variant modeling in SysML can be applied to a broad range of dissimilar systems, covering the entire domain of ground vehicles, in single reference architecture model. Traditionally, a system’s structure is decomposed into subsystems and components. However, this method is found to be ineffective when modeling variants that are functionally similar but structurally different. We propose to address this challenge by first decomposing the system not only by subsystem but also by high-level function. This pattern is particularly useful for situations where two variants perform the same function, but one variant performs the function using one subsystem, whereas the other variant performs the same function using one or more different subsystems.