Search Results

In-cylinder pressure measurement is an important tool in internal combustion engine research and development for combustion, cycle performance, and knock analysis in spark-ignition engines. In a typical laboratory setup, a sub crank angle resolved (typically between 0.1o and 0.5o) optical encoder is installed on the engine crankshaft, and a piezoelectric pressure transducer is installed in the engine cylinder. The charge signal produced by the transducer due to changes in cylinder pressure during the engine cycle is converted to voltage by a charge amplifier, and this analog voltage is read by a high-speed data acquisition (DAQ) system at each encoder trigger pulse. The high speed of engine operation and the need to collect hundreds of engine cycles for appropriate cycle-averaging requires significant processor speed and memory, making typical data acquisition systems very expensive.

Engineering design-decisions often involve many attributes which can differ in the levels of their importance to the decision maker (DM), while also exhibiting complex statistical relationships. Learning a decision-making policy which accurately represents the DM’s actions has long been the goal of decision analysts. To circumvent elicitation and modeling issues, this process is often oversimplified in how many factors are considered and how complicated the relationships considered between them are. Without these simplifications, the classical lottery-based preference elicitation is overly expensive, and the responses degrade rapidly in quality as the number of attributes increase. In this paper, we investigate the ability of deep preference machine learning to model high-dimensional decision-making policies utilizing rankings elicited from decision makers.

In the dense fabric of urban areas, electric scooters have rapidly become a preferred mode of transportation. As they cater to modern mobility demands, they present significant safety challenges, especially when interacting with pedestrians. In general, e-scooters are suggested to be ridden in bike lanes/sidewalks or share the road with cars at the maximum speed of about 15-20 mph, which is more flexible and much faster than pedestrians and bicyclists. Accurate prediction of pedestrian movement, coupled with assistant motion control of scooters, is essential in minimizing collision risks and seamlessly integrating scooters in areas dense with pedestrians. Addressing these safety concerns, our research introduces a novel e-Scooter collision avoidance system (eCAS) with a method for predicting pedestrian trajectories, employing an advanced Long short-term memory (LSTM) network integrated with a state refinement module.

Vacuum suction cups are used as transforming handles in stamping lines, which are essential in developing automation and mechanization. However, the vacuum suction cup will crack due to fatigue or long-term operation or installation angle, which directly affects production productivity and safety. The better design will help increase the cups' service life. If the location of stress concentration can be predicted, this can prevent the occurrence of cracks in advance and effectively increase the service life. However, the traditional strain measurement technology cannot meet the requirements of tracking large-field stains and precise point tracking simultaneously in the same area, especially for stacking or narrow parts of the suction cups. The application must allow multiple measurements of hidden component strain information in different fields of view, which would add cost.

Digital shearography has many advantages, such as full-field, non-contact, high sensitivity, and good robustness. It was widely used to measure the deformation and strain of materials, also to the application of nondestructive testing (NDT). However, most digital sherography applications can only work in one field of view per measurement, and some small defects may not be detected as a result. Multiple measurements of different fields of view are needed to solve this issue, which will increase the measurement time and cost. The difficulty in performing multiple measurements may also increase for cases where the loading is not repeatable. Therefore, a system capable of measuring dual fields of view at the same time is necessary. The carrier frequency spatial phase shift method may be a good candidate to reach this goal because it can simultaneously record phase information of multiple images, e.g. two speckle interferograms with different fields of view.

Controller Area Network (CAN), the de facto standard for in-vehicle networks, has insufficient security features and thus is inherently vulnerable to various attacks. To protect CAN bus from attacks, intrusion detection systems (IDSs) based on advanced deep learning methods, such as Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), have been proposed to detect intrusions. However, those models generally introduce high latency, require considerable memory space, and often result in high energy consumption. To accelerate intrusion detection and also reduce memory requests, we exploit the use of Binarized Neural Network (BNN) and hardware-based acceleration for intrusion detection in in-vehicle networks. As BNN uses binary values for activations and weights rather than full precision values, it usually results in faster computation, smaller memory cost, and lower energy consumption than full precision models.

Engineering practice routinely involves decision making under uncertainty. Much of this decision making entails reconciling multiple pieces of information to form a suitable model of uncertainty. As more information is collected, one expectedly makes better and better decisions. However, conditional probability assessments made by human decision makers, as new information arrives does not always follow expected trends and instead exhibits inconsistencies. Understanding them is necessary for a better modeling of the cognitive processes taking place in their mind, whether it be the designer or the end-user. Doing so can result in better products and product features. Quantum probability has been used in the literature to explain many commonly observed deviations from the classical probability such as: question order effect, response replicability effect, Machina and Ellsberg paradoxes and the effect of positive and negative interference between events.

Digital Image Correlation (DIC) technology is a powerful tool in the field of experimental mechanics to obtain the full-field deformation/strain information of an object. It has been rapidly applied in industry in recent years. However, for the large-angle full-field strain measurement of small-sized cylindrical objects, it’s still a challenge to the DIC accurate measurement due to its small size and curved surface. In this paper, a measurement method based on the multi-camera DIC system is proposed to study the compressive performance of small-sized cylindrical materials. Three cameras form two stereo DIC measurement systems (1 and 2 cameras, and 2 and 3 cameras), each of which measures a part of the object. By calibrating three cameras at the same time, two stereos DIC coordinate systems can be unified to one coordinate system. Then match the two sets of DIC measurement data together to achieve large-angle measurement of the cylindrical surface.

In the automotive industry, a Malfunction Indicator Light (MIL) is commonly employed to signify a failure or error in a vehicle system. To identify the root cause that has triggered a particular fault, a technician or engineer will typically run diagnostic tests and analyses. This type of analysis can take a significant amount of time and resources at the cost of customer satisfaction and perceived quality. Predicting an impending error allows for preventative measures or actions which might mitigate the effects of the error. Modern vehicles generate data in the form of sensor readings accessible through the vehicle’s Controller Area Network (CAN). Such data is generally too extensive to aid in analysis and decision making unless machine learning-based methods are used. This paper proposes a method utilizing a recurrent neural network (RNN) to predict an impending fault before it occurs through the use of CAN data.

This work presents a method to estimate the human body orientation using 2D images from a person view; the challenge comes from the variety of human body poses and appearances. The method utilizes OpenPose neural network as a human pose detector module and depth sensing module. The modules work together to extract the body orientation from 2D stereo images. OpenPose is proven to be efficient in detecting human body joints, defined by COCO dataset, OpenPose can detect the visible body joints without being affected by backgrounds or other challenging factors. Adding the depth data for each point can produce rich information to the process of 3D construction for the detected humans. This 3D point’s setup can tell more about the body orientation and walking direction for example. The depth module used in this work is the ZED camera stereo system which uses CUDA for high performance depth computation.

Expanding upon the authors’ previous work which utilized a GT-Power model of the Cooperative Fuels Research (CFR) engine under Research Octane Number (RON) conditions, this work defines the boundary conditions of the CFR engine under Motored Octane Number (MON) test conditions. The GT-Power model was validated against experimental CFR engine data for primary reference fuel (PRF) blends between 60 and 100 under standard MON conditions, defining the full range of interest of MON for gasoline-type fuels. The CFR engine model utilizes a predictive turbulent flame propagation sub-model, and a chemical kinetic solver for the end-gas chemistry. The validation was performed simultaneously for thermodynamic and chemical kinetic parameters to match in-cylinder pressure conditions, burn rate, and knock point prediction with experimental data, requiring only minor modifications to the flame propagation model from previous model iterations.

Research in any field of study requires analysis and comparisons or real-time predictions to extract useful information. To prove that the results have practical potential, various filtering techniques and methodologies should be designed and implemented. Filters being a class of signal processing helps innovate new technologies with various kinds of outcomes, using filters there are always various methods to solve a problem. Considering the current COVID-19 situation, researchers are working on sequencing the novel coronavirus and the genomes of people afflicted with COVID-19 using CPUs and GPUs along with various filtering techniques. In this paper we are using a method of filter implementation to collect raw heart rate data samples from fingertip and ear lobe and process those results on CPUs and GPUs. Our method of implementation to collect raw heart rate data is using a photoplethysmography method.

A new method is demonstrated for rating the “severity” of a hypoid gear set duty cycle (revolutions at torque) using the intercept of T-N curve to support gearset selection and sizing decision across vehicle programs. Historically, it has been customary to compute a cumulative damage (using Miner's Rule) for a rotating component duty cycle given a T-N curve slope and intercept for the component and failure mode of interest. The slope and intercept of a T-N curve is often proprietary to the axle manufacturer and are not published. Therefore, for upfront sizing and selection purposes representative T-N properties are used to assess relative component duty cycle severity via cumulative damage (non-dimensional quantity). A similar duty cycle severity rating can also be achieved by computing the intercept of the T-N curve instead of cumulative damage, which is the focus of this study.

This work presents a method for estimating human body orientation using a combination of convolutional neural network (CNN) and stereo camera in real time. The approach uses the CNN model to predict certain human body keypoints then transforms these points into a 3D space using the stereo vision system to estimate the body orientations. The CNN module is trained to estimate the shoulders, the neck and the nose positions, detecting of three points is required to confirm human detection and provided enough data to translate the points into 3D space.

An existing challenge in current state-of-the-art autonomous vehicles is the process of safely transferring control from autonomous driving mode to manual mode because the driver may be distracted with secondary tasks. Such distractions may impair a driver’s situational awareness of the driving environment which will lead to fatal outcomes during a handover. Current state-of-the-art vehicles notify a user of an imminent handover via auditory, visual, and physical alerts but are unable to improve a driver’s situational awareness before a handover is executed. The overall goal of our research team is to address the challenge of providing a driver with relevant information to regain situational awareness of the driving task. In this paper, we introduce a novel approach to estimating a driver’s visual focus of attention using a 2D RGB camera as input to a Multi-Input Convolutional Neural Network with shared weights. The system was validated in a realistic driving scenario.

Metal panels are comprehensively applied in the automotive industry. A significant issue with metal panels is the deflection when moving in the press line of the stamping process. Unpredictable deflection could result in the cut off of the press line. To control the deflection in a safe zone, finite element tools are used to simulate the panel transform process. However, the simulation requires experimental validation where conventional displacement measurement techniques could not satisfy the requirement of vast filed displacement and accuracy point tracking. In this study, multi-camera digital image correlation (DIC) systems have been developed to track the movement of panels during the press line of the stamping process. There are some advantages of applying the DIC system, including non-contact, full-field, high accuracy, and direct measurement techniques that provide the evaluation displacement of the metal panel and press line.

With the emergence of intelligent ground vehicles, an objective evaluation of vehicle mobility has become an even more challenging task. Vehicle mobility refers to the ability of a ground vehicle to traverse from one point to another, preferably in an optimal way. Numerous techniques exist for evaluating the mobility of vehicles on paved roads, both quantitatively and qualitatively, however, capabilities to evaluate their off-road performance remains limited. Whereas a vehicle’s off-road mobility may be significantly enhanced with intelligence, it also introduces many new variables into the decision making process that must be considered. In this paper, we present a decision analytic framework to accomplish this task. In our approach, a vehicle’s mobility is modeled using an operator’s preferences over multiple mobility attributes of concern. We also provide a method to analyze various operating scenarios including the ability to mitigate uncertainty in the vehicles inputs.

Lane line detection is a very critical element for Advanced Driver Assistance Systems (ADAS). Although, there has been significant amount of research dedicated to the detection and localization of lane lines in the past decade, there is still a gap in the robustness of the implemented systems. A major challenge to the existing lane line detection algorithms stems from coping with bad weather conditions (e.g. rain, snow, fog, haze, etc.). Snow offers an especially challenging environment, where lane marks and road boundaries are completely covered by snow. In these scenarios, on-board sensors such as cameras, LiDAR, and radars are of very limited benefit. In this research, the focus is on solving the problem of improving robustness of lane line detection in adverse weather conditions, especially snow. A framework is proposed that relies on using Vehicle-to-Infrastructure (V2I) communication to access reference images stored in the cloud.

Machine learning methods, such as decision trees and deep neural networks, are becoming increasingly important and useful for data analysis in various scientific fields including dynamics and control, signal processing, pattern recognition, fluid mechanics, and chemical synthesis, etc. For future engine design and performance optimization, there is an urgent need for a robust predictive model which could capture the major combustion properties such as autoignition and flame propagation of multicomponent fuels under a wide range of engine operating conditions, without massive experimental measurement or computational efforts. It will be shown that these long-held limitations and challenges related to complex fuel combustion and engine research could be readily solved by implementing machine learning methods.

Autonomous vehicles might one day be able to implement privacy preserving driving patterns which humans may find too difficult to implement. In order to measure the difference between location privacy achieved by humans versus location privacy achieved by autonomous vehicles, this paper measures privacy as trajectory anonymity, as opposed to single location privacy or continuous privacy. This paper evaluates how trajectory privacy for randomized driving patterns could be twice as effective for autonomous vehicles using diverted paths compared to Google Map API generated shortest paths. The result shows vehicles mobility patterns could impact trajectory and location privacy. Moreover, the results show that the proposed metric outperforms both K-anonymity and KDT-anonymity.