Browse Topic: Neural networks

Items (391)

Study on the Optimal Pre-Chamber Geometry for Active Pre-Chamber Gas Engines

2024-32-002304/18/2025

In a pre-chamber engine, fuel in the main-chamber is ignited and combusted by the combustion gas injected from the pre-chamber. Therefore, further fuel dilution is possible and thermal efficiency can be also improved. However, adding a pre-chamber to an engine increases the number of design parameters which have a significant impact on the main combustion and the exhaust gas. Then, in this study, the optimum geometry of the pre-chamber in an active pre-chamber gas engine was investigated. The considered parameters were the volume of pre-chamber, the diameter of a nozzle hole, and the number of nozzle holes. 18 types of pre-chambers with different geometries were prepared. Using these pre-chambers, engine experiments under steady conditions were conducted while changing the conditions such as engine speeds, mean indicated pressure and air excess ratio. Based on the experimental data, neural network models were constructed that predict thermal efficiency, NOx and CO emissions from the

Yasuda, Kotaro, Yamasaki, Yudai, Sako, Takahiro, Takashima, Yoshitane, Suzuki, Kenta

jms_bp_2 (Current)04/08/2025

Automated Vehicle Safety Consortium

Biohybrid Robotic Hand Research Platform

TBMG-5215912/01/2024

Brain-machine interfaces enable direct communication between a brain’s electrical activity and an external device such as a computer or a robotic limb that allows people to control machines using their thoughts. Researchers have developed a novel biohybrid neuroprosthetic research platform comprised of a dexterous artificial hand electrically interfaced with biological neural networks.

ANN Based Surrogate Model for Fillet Stress Prediction of Automotive Crankshafts

2024-26-028201/16/2024

Designing of automotive crankshaft is a time-consuming process and often involves several iterations of crankshaft designs being validated by calculations till a safe design is conceptualized. In the initial phases of design process, Finite Element Analysis (FEA) methods are conventionally used for design validation of crankshaft by stress and fatigue calculations. Setting up of FEA models to find crankshaft fillet stress is laborious and often the design is required to undergo a few iterations of changes before it can be deemed safe for the next phases of development. This study aims to implement an Artificial Neural Network (ANN) prediction model to predict crankshaft fillet stress under a 3-point bending scenario and evaluates its performance. The objective of this study was to develop a Machine learning prediction model based on Artificial Neural Networks that predicts maximum von mises stress at Crankshaft fillets region. A dataset containing previous FEA stress data was used to

Thokalath, Jaison Jacob

Recurrent Neural Network Model for On-Board Estimation of the Side-Slip Angle in a Four-Wheel Drive and Steering Vehicle

09/23/2023

Abstract A valuable quantity for analyzing the lateral dynamics of road vehicles is the side-slip angle, that is, the angle between the vehicle’s longitudinal axis and its speed direction. A reliable real-time side-slip angle value enables several features, such as stability controls, identification of understeer and oversteer conditions, estimation of lateral forces during cornering, or tire grip and wear estimation. Since the direct measurement of this variable can only be done with complex and expensive devices, it is worth trying to estimate it through virtual sensors based on mathematical models. This article illustrates a methodology for real-time on-board estimation of the side-slip angle through a machine learning model (SSE—side-slip estimator). It exploits a recurrent neural network trained and tested via on-road experimental data acquisition. In particular, the machine learning model only uses input signals from a standard road car sensor configuration. The model

Giuliacci, Tiziano Alberto, Ballesio, Stefano, Fainello, Marco, Mair, Ulrich, King, Julian

Recurrent Neural Network Model for On-Board Estimation of the Side-Slip Angle in a Four-Wheel Drive and Steering Vehicle

15-17-01-0003-TEST-REC09/23/2023

Abstract A valuable quantity for analyzing the lateral dynamics of road vehicles is the side-slip angle, that is, the angle between the vehicle’s longitudinal axis and its speed direction. A reliable real-time side-slip angle value enables several features, such as stability controls, identification of understeer and oversteer conditions, estimation of lateral forces during cornering, or tire grip and wear estimation. Since the direct measurement of this variable can only be done with complex and expensive devices, it is worth trying to estimate it through virtual sensors based on mathematical models. This article illustrates a methodology for real-time on-board estimation of the side-slip angle through a machine learning model (SSE—side-slip estimator). It exploits a recurrent neural network trained and tested via on-road experimental data acquisition. In particular, the machine learning model only uses input signals from a standard road car sensor configuration. The model

Giuliacci, Tiziano Alberto, Ballesio, Stefano, Fainello, Marco, Mair, Ulrich, King, Julian

Applications of Neural Networks to Metallic Flexor Geometry Optimization of Flat Wipers

09/09/2023

Abstract In recent years, demands of flat wipers have rapidly increased in the vehicle industry due to their simpler structure compared to the conventional wipers. Procedures for evaluating the appropriate metallic flexor geometry, which is one of the major components of the flat wiper, were proposed in the authors’ previous study. However, the computational cost of the aforementioned procedures seems to be unaffordable to the industry. The discrete Winkler model regarding the flexor as the Euler–Bernoulli beam is established as the mathematical model in this study to simulate a flexor compressed against a surface at various wiping angles. The deflection of the beam is solved using a finite difference method, and the calculated contact pressure distributions agree fairly with those based on the corresponding finite element model. Flexor designs are paired with various windshield surfaces to accumulate a sufficiently large simulation database based on the mathematical model. An

Chu, Yi-Tzu, Huang, Ting-Chuan, Liao, Kuo-Chi

Categorizing Simulation Models Using Convolutional Neural Networks

2023-01-121706/26/2023

Engineers come across frequent problems solved using artificial intelligence. Tasks of neural networks are often considered solving problems in such an intelligent manner. Whether as an optimization problem or development tool, neural networks help engineers and create a more efficient approach. Furthermore, information is stored in a unique data structure, during a project cycle. Engineers often face a time-consuming search to get an understanding of the project itself. This process of understanding the data needs to be repeated by every engineer, if not further documented. E.g. searching for a past project with similar parameters can be a big task, regarding huge databases. Worker fluctuation can increase the cost of this process as well. To address this issue, neural networks make a significant contribution. The main aspect of this paper is to add meta data to the core files of the project. With such labels it is possible to get access to the content of the model files of an engine

Grbavac, Andrija, Angerbauer, Martin, Grill, Michael, Kulzer, André

Application of AI/ML in Hydrobush Tuning to Enhance Overall Value Proposition

2025-01-013205/05/2023

In this intense market of the automotive industry, optimization of vehicle components for superior performance and customer satisfaction is paramount. Hydrobushes play an integral role in achieving this within vehicle suspension systems by absorbing vibrations and improving ride comfort. The methods traditionally used for tuning these components are not only time-consuming but also heavily reliant on extensive empirical testing. This paper explores the ever-growing field of artificial intelligence (AI) and machine learning (ML) in the hydrobush tuning process. Leveraging AI/ML algorithms such as random forest, artificial neural network, and logistic regression to analyse vast datasets, identify patterns, and predict optimal configurations more efficiently than conventional methods. The stated approach accelerates the tuning process while also enhancing the precision of the outcomes, leading to superior ride quality and durability. Thus, the integration of AI/ML techniques in hydrobush

Hazra, Sandip, Khan, Arkadip Amitava

Mode Shape Identification Using Graph Neural Networks for Vehicle Structure Design

2025-01-013105/05/2023

This paper introduces a novel, automated approach for identifying and classifying full vehicle mode shapes using Graph Neural Networks (GNNs), a deep learning model for graph-structured data. Mode shape identification and naming refers to classifying deformation patterns in structures vibrating at natural frequencies with systematic naming based on the movement or deformation type. Many times, these mode shapes are named based on the type of movement or deformation involved. The systematic naming of mode shapes and their frequencies is essential for understanding structural dynamics and “Modal Alignment” or “Modal Separation” charts used in NVH analysis. Current methods are manual, time-consuming, and rely on expert judgment. The integration of GNNs into mode shape classification represents a significant advancement in vehicle modal identification and structure design. Results demonstrate that GNNs offer superior accuracy and efficiency compared to current labor intensive manual modes

Tohmuang, Sitthichart, Swayze, James L., Fard, Mohammad, Fayek PhD, Haytham, Marzocca, Piergiovanni, Bhide, Sanjay, Huber, John

Investigations on the Effect of Injection Pressure and Timing on Noise and Vibration Characteristics of a Biodiesel Fueled Genset Engine and Artificial Neural Network Architecture

2022-28-049911/18/2022

In this present investigation, combustion, noise and vibration characteristics were studied in a direct injection compression ignition engine by blending of 20% Pongamia biodiesel with neat diesel. Experiments were conducted with different fuel injection pressures and injection timings. The brake thermal efficiency of the biodiesel blend found increased to 1.75% when the fuel was injected at 200 bar and 23° before top dead center. Higher engine vibration of 5.5% was observed in Y-axis at the fuel injection pressure of 250 bar with the injection timing of 25° bTDC when comparing to the standard operating condition of the engine. The combustion noise level found decreased to 4% when the fuel injection time increased from 21 to 23° bTDC. From Artificial Neural Network analysis, two metrics, such as Coefficient of correlation (R) and Mean Square Error (MSE), were found as 0.99 and 12.6, respectively. The developed artificial neural network model predicts and optimizes the parameters used

Selvan, V. Arul Mozhi, G, Arumugam

3D Printing Materials with Tunable Mechanical Properties

TBMG-4691211/01/2022

Teaching Drones How to Navigate Real-World Weather

TBMG-4649509/01/2022

Right now, drones are either flown under controlled conditions, with no wind, or are operated by humans using remote controls. However, for drones to autonomously perform necessary but quotidian tasks, such as delivering packages or airlifting injured drivers from a traffic accident, drones must be able to adapt to wind conditions in real time.

A Novel Cloud-Based Additive Manufacturing Technique for Semiconductor Chip Casings

05-15-04-002508/02/2022

The demand for contactless, rapid manufacturing has increased over the years, especially during the COVID-19 pandemic. Additive manufacturing (AM), a type of rapid manufacturing, is a computer-based system that precisely manufactures products. It proves to be a faster, cheaper, and more efficient production system when integrated with cloud-based manufacturing (CBM). Similarly, the need for semiconductors has grown exponentially over the last five years. Several companies could not keep up with the increasing demand for many reasons. One of the main reasons is the lack of a workforce due to the COVID-19 protocols. This article proposes a novel technique to manufacture semiconductor chips in a fast-paced manner. An algorithm is integrated with cloud, machine vision, sensors, and email access to monitor with live feedback and correct the manufacturing in case of an anomaly. Several real-time data such as performance graphs, time left, and other current information were sent back to the

Viswanath, Shreya, Siddharth, S., Subramanian, Jeyanthi

AI deep learning-based bird detection for UAM and aircraft

2022-26-122805/26/2022

Abstract - Safety of flight during takeoff, flight and landing is one of the important considerations for airborne vehicles. Bird strike is defined as a collision between a bird and aircraft during takeoff, flight and landing of airborne vehicles. When bird strike happens, it can result in damage to aircraft structure and loss of thrust which impacts safety of flight. The nature of aircraft damage can differ depending on the size of the aircraft. For smaller aircrafts, it can result in penetration into flight deck windscreens or damage to control surfaces. In case of larger aircraft, it can result in hazardous effects because of engine ingestion. Many deep learning algorithms have evolved for object detection and classification in the recent years. You Only Look Once [YOLO] and single shot detector [SSD] are top two state of art object detection algorithms that are extensively used in real world applications. The Xilinx® Deep Learning Processor Unit (DPU) is a programmable engine

USE OF AI BASED VISUAL INSPCTION FOR AEROSPACE PARTS INSPECTION AND MANUFACTURING APPLICATION Govind Yadav Siva Bavisetti Honeywell Technology Solutions Bangalore, Karnataka, India

2022-26-122505/26/2022

ABSTRACT Aircraft manufacturing processes typically include one or more steps where the product is visually inspected for defects. Typically, visual inspection is a highly manual process that can be time consuming and prone to errors. Over the years, rule-based visual inspection machines have also emerged. However, each approach has drawbacks. Manual inspection requires the presence of a person, an inspector who performs assessment of the entity under question and passes judgement on it according to some training or previous knowledge. No equipment is required except the naked eye of the trained inspector. Visual inspection errors in manufacturing take one of two forms — missing an existing defect or incorrectly identifying a defect that does not exist (false positive). Misses tend to occur much more frequently than false alarms). Misses can lead to loss in quality, while false positives can cause unnecessary production costs and overall wastage. This paper studies the limitations of

Yadav, Govind, Bavisetti, Siva

A Framework for Teaching Safety Critical Control and Artificially Intelligent Systems to Undergrads

2022-26-122305/26/2022

There is a need to teach systems to the undergrad and graduate levels in colleges. Efforts are being made by industry, academia, and professional societies to join hands to bridge the gap. The students require examples to get their “hands dirty” on an actual system. Many of the aerospace examples are not available in the open domain as they are considered proprietary. The world is going toward Artificial Intelligence. Though AI has been used in aerospace industry for optimization for quite some time, certification of AI systems in the safety critical control domain are still not very popular. Experiments on such system are required to clearly understand the testing and safety aspects of deploying such systems. This paper looks at an environmental control system which has enough complexity to merit an industry level attention but is again simple enough to be understood by the students. This is a physics model. Simulation runs from this physics model is used to train a Neural Network

Jeppu, Yogananda, Raman, Ramakrishnan

An Autonomous Ground Based Multi Sensor Data Fused TDR Prediction and Monitoring System

2022-26-122705/26/2022

According to a study conducted by IATA, EASA and ICAO, runway excursions are common type of incidents that leads to major accidents. Runway excursion predominantly occur due to incorrect utilization of the touchdown zone (TDZ). A ground-based multi-sensor data fused touchdown region (TDR) prediction and monitoring system is proposed in this paper. This system autonomously analyses real-time environmental data with remote sensing capability around the runway site, gathers real-time information on flight critical parameters of approaching aircraft through an ADS-B Out data link, predicts ideal TDR over the runway TDZ for a particular aircraft, and estimates the expected contemporaneous TDR based on the monitored current approach profile. If the ideal and expected TDR do not overlap, there is a chance for an excursion. In such cases, the flight parameter that needs corrective action is determined and notified to the pilot. In order to analyze dynamic environmental data, mathematical

Nagonda, Ajay Nagarajan, Manjunath, Karthik

Navigating Deep Learning to Improve ADAS

22AVEP04_0404/01/2022

How edge computing drives real-time decision making in AI for smarter, safer automated vehicles. Automotive ‘big data’ is here. The immense scope of data generated by automated or ADAS-integrated vehicles spans the five SAE levels of autonomous driving, with reliance on high-resolution cameras, radar, lidar, ultrasonic sensors, GPS and other sensors for vehicles to see or perceive their surroundings. Ultimately, this sensory information - massive amounts of data - is used to navigate, avoid obstacles and read road markers necessary for safe driving. Artificial intelligence (AI) is at the heart of these operations, grounded in software algorithms and fueled by deep-learning training and deep-learning inference models that are essential to faultless performance. Enabling these vital and instantaneous processes requires AI algorithms to be trained and then deployed on-vehicle. It's a process that has developers tapping into both sophisticated software design and smart hardware strategies

Seetoo, Dustin, SintzUSER, Jeneane

KDepthNet: Mono-Camera based Depth Estimation for Autonomous Driving

2022-01-007903/29/2022

Object avoidance for autonomous driving is a vital factor in safe driving. Having the accurate depth of each barrier in the scene can contribute to obstacle prevention. In recent years, precise depth estimation systems can be attributed to notable advances in Deep Neural Networks and hardware facilities/equipment. There are several depth estimation methods for autonomous vehicles which usually utilize lasers, structured light and other reflections on the object surface to capture depth point clouds, complete surface modelling and estimate scene depth maps. However, estimating precise depth maps is still challenging. Indeed, it usually requires heavy time cost, abundant manpower and huge computing devices. Consequently, image-based depth calculation has turned into the mainstream of study and can be applied for a broad range of applications. A vast majority of camera depth estimation methods intend to determine the depth map of the whole input image using binocular cameras or a 3D

Tavakolian, Niloofar, Fekri, Pedram, Zadeh, Mehrdad, Dargahi, Javad

Research on Brake Comfort Based on Brake-by-Wire system Control

2022-01-111303/29/2022

The vehicle will produce certain shock and vibration during the braking process, which will affect the driving experience of the driver. Aiming at the problems of pitch vibration, longitudinal vibration and shock during the braking process, this paper proposes a planning and following control method for target longitudinal acceleration in post-braking phase, and designs control trigger and termination strategies. Target longitudinal acceleration planning takes minimizing longitudinal shock as the design goal. The following control takes the brake pressure as the control object, and adopts the "feedforward +PID feedback" method to follow the target longitudinal acceleration. Based on Carsim and Simulink software, the simulation model of straight-line braking is established. In this model, the control target is longitudinal acceleration and the control object is four-wheel braking pressure. The results of straight-line braking simulation show that the proposed longitudinal acceleration

Xiong, Wei, Wang, Weiwei, Jiang, Kaihong, Hou, Liming, Shangguan, Wen-Bin

Anticipation-Based Autonomous Platoon Control Strategy with Minimum Parameter Learning Adaptive Radial Basis Function Neural Network (MPL-ARBF-NN) Sliding Mode Control

2022-01-034803/29/2022

Human-driver is stochastic in nature and usually difficult to predict. In car following scene, where an autonomous vehicle platoon follows a human-driven vehicle having a complex acceleration behavior therefore needs greater attention for smooth traffic flow. To guarantee a string stable platoon operation, an efficient vehicle controller must be developed for fast convergence, adaptation simplicity, ease of design and real-time performance. The research proposed an adaptive fuzzy system-based compound variable exponential power reaching law (CVEPRL) sliding mode control in a backstepping design framework. The fuzzy system employs an efficient adaptation to minimize computational burden in real-time operation of the platoon. The ease of controller development arise from the backstepping control design and robustness of the controller can be inherited from sliding mode control (SMC). The autonomous vehicle can be modeled as a third order nonlinear plant with a varying disturbance term

Negash, Natnael M., Yang, James

Optimization of Computer Vision Software Models for Deployment to Performance Constrained Embedded Processors

2022-01-017903/29/2022

Computer vision (CV), a form of artificial intelligence (AI), is a foundational technology within the automotive industry for an increasing number of applications including active safety, motion control, and driver distraction monitoring. State-of-the-art CV models often rely on the use of Deep Neural Networks (DNNs) to achieve high levels of accuracy. While necessary for their accuracy, DNNs are computationally complex. For example, when compared to other AI model architectures, they have a large memory footprint and must perform a high number of operations to create an output or prediction. To meet performance goals in the face of such constraints, high performance processors such as Graphics Processing Units (GPUs) are typically required to run CV models on-board automobiles, creating a major hurdle to the deployment of CV applications. This paper proposes and analyzes a method for optimizing and compressing DNN-based CV models to enable their deployment to lower performance

Rocci, Benjamin M.

Dynamic Behavior Prediction of Magnetorheological Fluid Dampers using Neural Networks

2022-01-032803/29/2022

Parametric and non-parametric models have been studied to predict the behavior of magnetorheological (MR) fluid dampers by a lot of researchers due to their nonlinear dynamics. In this paper, the direct and inverse dynamic identification for MR fluid dampers using recurrent neural networks are investigated to demonstrate more accurate and efficient model. The effect of neural networks construction on the prediction quality of dynamic performance of MR damper is introduced in details. The trained direct identification neural network model can be used to predict the damping force of the MR fluid damper on line and the inverse dynamic neural network model can be used to generate the command voltage applied to the damper coil through supervised learning. The architectures and the learning techniques of direct and inverse neural network models for MR fluid dampers are introduced and simulation results are discussed. Finally, the trained neural network models are used to predict the damping

abd elwahed, A., Monieb, H. A., Metered, H.

Vehicle Feature Recognition Method Based on Image Semantic Segmentation

2022-01-016503/29/2022

In the process of truck overload and over-limit detection, it is necessary to detect the characteristics of the vehicle's size, type, and wheel number, etc. In addition, in some vehicle visual load recognition systems, the vehicle load can be calculated by detecting the vibration frequency of specific parts of vehicle or the change in the length of the suspension during the vehicle's forward process. Therefore, it is very important to quickly and accurately identify vehicle features through the camera. This paper proposes a vehicle feature recognition method based on image semantic segmentation and Python, which can identify the length, height, number of wheels and vibration frequency at specific parts of the vehicle based on the vehicle driving video captured by the roadside camera. The process of vehicle recognition is as follows: First, build a convolutional neural network based on the image semantic segmentation model SegNet and use the data set to train it, and use the Python

Wang, Chu, Ding, Shousheng

The Virtual Boosted DISI Engine Model Development Based on Artificial Neural Networks

2022-01-045603/29/2022

To efficiently reduce the required experimental data and improve the prediction accuracy, a virtual engine mod-el has been built by integrating an artificial neural network (ANN) system consisting of multiple subnets with the principle component analysis (PCA) method. The PCA method could preprocess the training data and leads to a more efficient training process. By coupling with 1-D GT power platform, this model has been adopted to predict the combustion phases (including CA10, CA50 and CA90), exhaust gas temperature, brake specific fuel consumption rate (be) and engine emissions such as un-burnt hydrocarbon (UBHC), NOx and CO. The results are then compared with the experimental data from around 5000 operating points of a boosted DISI engine running at universal performance map and conditions with various valve timing configurations. The mean abso-lute errors in combustion phases and exhausted gas temperature are less than 1.0 °CA and 10.1 K, respectively. The average relative errors

Chen, Ceyuan, Wu, Jian, Wei, Jingsi, Xu, Hanjun

Road Crossing Assistance Method Using Object Detection Based on Deep Learning

2022-01-016903/29/2022

This paper describes a method for assisting pedestrians to cross a road. As motorization develops, pedestrian protection techniques are becoming more and more important. Advanced driving assistance systems (ADAS) are improving rapidly to provide even greater safety. However, since the accident risk of pedestrians remains high, the development of an advanced walking assistance system for pedestrian protection may be an effective means of reducing pedestrian accidents. Crossing a road is one of the highest risk events, and is a complex phenomenon that consists of many dynamically changing elements such as vehicles, traffic signals, bicycles, and the like. A road crossing assistance system requires three items: real-time situational recognition, a robust decision-making function, and reliable information transmission. Edge devices equipped with autonomous systems are one means of achieving these requirements. Situational recognition when crossing a road must identify the pedestrian

Kawamura, Hiroaki, Shintani, Kohei, Mima, Hiroki, Taniguchi, Mashio

Databased modeling: An AI Toolchain for the development process of combustion engines

2022-01-018803/29/2022

Predictive physical modeling is an established method used in the development process for any automotive component or system. While accurate predictions can be issued after tuning model parameters, long computation times can be expected depending on the complexity of the model at hand. As the requirements of the components/systems to be developed continuously increase, new optimization approaches are constantly being applied to solve multidimensional objectives and resulting conflicts optimally. Unfortunately, some of those approaches are deemed not feasible, as the computational times of the required single predictions using conventional simulation models are too high. Therefore, it is proposed to use data-based models such as trained neural networks instead of physical models to address this issue. Previous efforts have failed due to a weak database and the resulting poor predictive ability of data-based models. This paper introduces an AI Toolchain used for data-based modeling of

Milojević, Sasa, Rether, Dominik, Bodza, Sebasian, Grill, Michael, Angerbauer, Michael, Bargende, Michael, Cimniak, Valerian

Identification of Road Adhesion Coefficient Based on Neural Network and Kalman Filter

2022-01-110303/29/2022

As an important input parameter of intelligent vehicle active safety technology, road adhesion coefficient is of great significance in autonomous collision avoidance, emergency braking and collision avoidance, and variable adhesion road motion control. Traditional recognition methods based on vehicle dynamics require large data volume and low solution accuracy, while recognition methods based on optical sensors have strict requirements on the working environment, poor adaptability under complex working conditions, and high cost. This paper proposes an adhesion coefficient recognition method based on Elman neural network and Kalman filter. By establishing a four-degree-of-freedom vehicle dynamics model, the dynamic parameters such as yaw rate, longitudinal velocity, lateral velocity, and angular velocity of each wheel related to the road adhesion coefficient are analyzed as the input of the neural network model. Carsim/Simulink co-simulation is used to establish a data set, the Kalman

A prediction model of octane number loss based on neural network

2022-01-018103/29/2022

The octane number is the most important indicator of motor petrol, and the petrol refining process is one of the important links in petrol production. The prediction of the octane loss number of petrol during the refining process is helpful to the improvement of petrol refining process and the processing of petrol. The traditional octane number prediction method relied on physical and chemical properties, and did not fully consider the high nonlinearity and strong coupling relationship of the petrol refining process.There is a lack of data-driven octane loss models. This paper studies the construction of the octane loss model in the petrol refining process. The main innovation is to use frequency statistics to summarize the operating range of data variables and preprocess the dataset, and use the variance selection method to filter out some invalid variables, then use the multi-dimensional association analysis to select 27 main operating variables in the next step. Finally, a neural

yang, Gang, Zhang, Sumin, Li, Shuai, Li, Wenju

A prediction method of flow characteristics of electronic expansion valve based on improved particle swarm optimization neural network algorithm

2022-01-020403/29/2022

Electronic expansion valve as a throttle element is widely used in heat pump systems and flow characteristics are its most important parameter. The flow characteristics of the electronic expansion valve (EXV) with a valve port diameter of 3mm are studied, when the refrigerant R134a is used as the working fluid. The main factors affecting the flow characteristics are researched by adopting the orthogonal experiment method, for example, inlet pressure, inlet subcool, outlet pressure and valve opening. In view of the complicated phase change of the refrigerant passing through the electronic expansion valve, it is difficult to model the flow characteristics accurately. Based on the measured experimental data, an artificial neural network method with weights and thresholds optimized by improved particle swarm algorithm is used to predict the flow characteristics of the valve. A comparative analysis is carried out with an artificial neural network method without the improved particle swarm

Liang, Gaoshuai, Li, Liping, Shangguan, Wen-Bin

An Efficient Machine Learning Algorithm for Valve Fault Detection

2022-01-018503/29/2022

Multi-level Miller-cycle Dynamic Skip Fire (mDSF) is a combustion engine technology that improves fuel efficiency by deciding on each cylinder-event whether to skip (deactivate) the cylinder, fire with low (Miller) charge, or fire with a high (Power) charge. In an engine with two intake and two exhaust valves, skipping the cylinder is accomplished by deactivating all valves, while firing with a reduced charge is accomplished by deactivating one of the intake valves. This new ability to modulate the charge level introduces new failure modes. The first is a failure to reactivate the single intake valve, which results in a desired high-fire having the air intake of a low-fire; the second is a failure to deactivate the single intake valve, which results in a low-fire having the air intake of a high-fire. Reliably detecting these two faults has proven challenging for classical techniques that use measured MAP and/or crank angle acceleration to identify characteristic features of the

Serrano, Joe, Ortiz-Soto, Elliott, Chen, S Kevin, Chien, Li-Chun, Joshi, Abhishek

Machine learning assisted analysis of heat transfer characteristics of a heavy duty natural gas engine

2022-01-057203/29/2022

Heat transfer affects engine performance, efficiency, and emissions. Hence, it was essential to investigate the heat transfer characteristics after converting a compression ignition engine to a spark ignition and natural gas. The heat transfer for each engine speed and load was calculated based on Woschni’s correlation. As the number of experiments that provided engine response was limited, an artificial neural network approach was proposed to predict the engine heat transfer characteristics and aid in the analysis. The good agreement between the experimental results and data predicted by the machine learning algorithm validated the predictive capability of the artificial neural network model. The results of the data-driven analyses suggested that the heat transfer increased as the spark timing advanced. Moreover, increasing the engine speed reduced the heat flux form the combustion chamber to the engine coolant. Further, leaning the mixture lowered the heat transfer rate to the

Liu, Jinlong, Dumitrescu, Cosmin, Ulishney, Christopher

Research on neural network-based prediction model for CNG engine intake charge

2022-01-018403/29/2022

Based on the sample data obtained from the bench test of a four-cylinder naturally aspirated CNG engine, three different neural networks, BP, SVM and GRNN, were used to develop the intake charge prediction model for the intake system of this engine, in which engine speed, intake manifold pressure and intake temperature, VVT angle and gas injection time were taken as input parameters and intake charge was used as output parameter. The comparative analysis of the experimental data and model prediction data showed that the mean absolute error (MAE) of BP model, SVM model, and GRNN model were 2.69, 5.13, and 8.11, and the root mean square error (MSE) were 3.53, 7.17, and 9.29, respectively. BP neural network has smaller prediction error and higher accuracy than SVM and GRNN neural network, which is more suitable for the prediction of the intake charge of this type of four-cylinder naturally aspirated CNG engine.

zhang, peng, Ni, Jimin, Shi, Xiuyong

Performance optimization of road noise system based on convolutional neural network

2022-01-037003/29/2022

Abstract: In view of the traditional road noise active control system being unable to adapt to different types of road excitations, a convolutional neural network method is designed to identify the road surface in advance, and the convergence factor of the road noise system algorithm is adaptively controlled by different road input to make the vehicle on different roads. To achieve the best noise reduction effect. This paper collects three typical road images, uses migration learning to train and test the network, and combines the recognition results with the road noise active control algorithm for joint simulation. The results show that the road surface can be recognized accurately by using convolutional neural network. According to different road types, by adjusting the algorithm convergence factor, the noise reduction effect of the vehicle can be effectively improved.

Shi, Chenlu, Zhu, Yugang, Sen, Zheng, Ge, Xu, lv, Xiao, Zihong, Ling

Natural rubber life estimation through an extreme learning machine

2022-01-029703/29/2022

In engineering applications, rubber isolators are subjected to continuous alternating loads, resulting in fatigue failure. Although some theoretical models are used for the fatigue life estimation of rubber materials, they do not comprehensively consider the influences of multiple factors. In the present study, a model based on the extreme learning machine (ELM) is established to estimate fatigue life of natural rubber (NR) specimens. The mechanical load (engineering strain peak), ambient temperature (23℃, 60℃ and 90℃) and shore hardness (N45 and N50) of NR specimens are used as the input variables while the measure average fatigue life as the output variable of the ELM. The regression results and predicted life distribution of the established ELM model are encouraging. For comparison, the back propagation neural network (BPNN) model and the support vector machine (SVM) model are also implemented. It is concluded that the ELM model is superior to the other two ML models in accuracy and

Performance prediction of Proton Exchange Membrane Hydrogen Fuel Cells using the GRU model

2022-01-083803/29/2022

In recent years, fuel cell vehicles have attracted more attention because of the advantages of no environmental pollution and high energy density, however, the cost and durability of fuel cells have been important factors limiting the rapid development of fuel cell vehicles. How to quickly predict the life of fuel cells has always been the emphasis and focus of the industry. Therefore, this paper mainly focuses on two sets of proton exchange membrane hydrogen fuel cell durability test data. In this paper, we establish a fuel cell life prediction model to carry out product prediction research, using Gated Recurrent Unit Neural Network (GRU-NN)——a variant of "Recurrent Neural Networks" (RNN). This article first divides the two sets of fuel cell durability test data into a training group and a verification group, and trains the established neural network model with the test data of the training group. The output of the model is the polarization curve (current-voltage curve) and voltage

Yao, Ruimeng, Gu, Rongxin, Zhong, Minghui, Li, Wenqi, Jiang, Changlong, Pan, Xiangmin

Dynamic durability prediction of fuel cells using Long short-term memory neural network

2022-01-083703/29/2022

Durability performance prediction is a critical issue in fuel cell research. During the demonstration operation of fuel cell commercial vehicles in China, this issue has attracted more attention. In this article, the long short-term memory neural network (LSTMNN), which is an improved recurrent neural network (RNN), and the demonstration operation data are used to establish the prediction model to predict the durability performance of fuel cell stack. Then, a model based on back-propagation neural network (BPNN) is established to be a control group. The demonstration operation data is divided into training group and validation group. The former is used to train the prediction model, and the latter is used to verify the validity and accuracy of the prediction model. The outputs of the prediction model, as the durability performance evaluation indexes of fuel cell, are the polarization curve (current-voltage curve) and the voltage decay curve (time-voltage curve). Moreover, mean absolute

Ma, Liying, Peng, Suhang, Zhong, Minghui, Li, Wenqi, Jiang, Changlong, Pan, Xiangmin

Estimation of Surface Temperature Distributions Across an Array of Lithium-Ion Battery Cells Using a Long Short-Term Memory Neural Network

2022-01-084503/29/2022

As electric vehicles are becoming increasingly popular and necessary for the future mobility needs of civilization, further effort is continually made to improve the efficiency, cost, and safety of the lithium-ion battery packs that power these vehicles. To facilitate these goals, this paper introduces one possible type of model to predict a distribution of surface temperatures for a lithium-ion battery pack: a long short-term memory (LSTM) neural network. The LSTM model is trained and validated with lithium-ion cells electrically connected to form a battery pack. Voltage, current, state of charge (SOC), and cell surface temperature from one array are used as inputs from a wide range of high and low temperature drive cycles. Additionally, a second model is considered where ambient temperature is added as an input to the LSTM model and the resulting predictions are compared to the original model. In summary the LSTM model can accurately characterize and predict a distribution of lithium

Campbell, Jeffrey, ZHU, DI, Cho, Gyouho

Identification of the Nonlinear Dynamic Behavior of Magnetorheological Fluid Dampers using Adaptive Neuro-Fuzzy Inference System

2022-01-032903/29/2022

Adaptive neuro-fuzzy inference system (ANFIS) technique has been developed and applied by numerous researchers as a very useful predictor for nonlinear systems. In this paper, non-parametric models have been investigated to predict the direct and inverse nonlinear dynamic behavior of magnetorheological (MR) fluid dampers using ANFIS technique to demonstrate more accurate and efficient models. The direct ANFIS model can be used to predict the damping force of the MR fluid damper and the inverse dynamic ANFIS model can be used to offer a suitable command voltage applied to the damper coil. The architectures and the learning details of the direct and inverse ANFIS models for MR fluid dampers are introduced and simulation results are discussed. The suggested ANFIS models are used to predict the damping force of the MR fluid damper accurately and precisely. Moreover, validation results for the ANFIS models are proposed and used to evaluate their performance. Validation results with several

abd elwahed, A., Monieb, H. A., Metered, H.

Prediction of Lane Change on the Expressway Based on Logistic Regression

2021-01-703212/15/2021

In autonomous driving system, lane change decision-making plays an important role as the front-end of lateral control. However, the current prediction methods of lane change are typically performed by using basic variables as the features of model without deep processing, which reduces the accuracy of the prediction. Therefore, we propose a binary logistic regression method to solve the lane change decision problem under expressway conditions, which treat quantified willingness and risk as the inputs. Firstly, we design Lane Changing Willingness function and Lane Changing Risk function with the minimum safety spacing theory and traffic environment factors. Secondly, a binary logistic regression method for predicting lane change behavior is proposed. Thirdly, we develop the driving simulation platform with low latency data collecting tools and design the experiments. After training the model with the experiment data, the proposed method predicts the lane change decision with 94.02

Sun, Weiqi, Bai, Jie, Huang, Libo, Chang, Liang, Dong, Lianfei, Luo, Zhengang

Fault Detection and Diagnosis of Engine Spark Plugs Using Deep Learning Techniques

03-15-04-002711/10/2021

Fault Detection and Diagnosis (FDD) is playing an increasingly important role in the automotive sector as it moves toward Advanced Technology Vehicles. Reducing the cost of sensory equipment to detect faults in Internal Combustion Engines (ICEs) has always been a common desire for automotive researchers. This article offers an Artificial Intelligence approach for detecting engine combustion faults related to spark plugs using existing sensors. The study investigates two deep learning models that are capable of learning different fault conditions from historical sensory data. The two customized models, one Long Short-Term Memory (LSTM) neural network and one Convolutional Neural Networks (CNN) model, are proposed to tackle this task. The LSTM model processes the filtered sensor data in time series, while the CNN model uses the frequency map that is novel in the learning-based engine diagnosis field. A comprehensive engine fault dataset is collected and includes a variety of operating

Huangfu, Yixin, Seddik, Essam, Habibi, Saeid, Wassyng, Alan, Tjong, Jimi

Estimation of End of Life of Lithium-Ion Battery Based on Artificial Neural Network and Machine Learning Techniques

2021-26-021809/22/2021

Various vehicle manufacturers are launching electric vehicles, which are more sustainable and environmentally friendly. The major component in electric vehicles is the battery, and its performance plays a vital role. Usually the end of life of a battery in the automobile sector is when the battery capacity reaches 80% of its maximum rated capacity. The capacity of a lithium-ion cell decreases with the number of cycles. So, a semi-empirical model is developed for estimating the maximum stored capacity at the end of each cycle. The parameters considered in the model explain the changes in battery internal structure, like capacity losses at different conditions. The capacity estimated using the semi-empirical model is further taken as the inputs for estimating capacity using the Artificial Neural Network (ANN) and Machine Learning (ML) techniques i.e. Linear Regression (LR), Gaussian Process Regression (GPR), Support Vector Machine methods (SVM). Artificial Neural Network model has been

Lahari, Kothala, Sharma, Minakshi, Baidya, Kapil

A Comparative Study of Recurrent Neural Network Architectures for Battery Voltage Prediction

2021-01-125209/21/2021

Electrification is the well-accepted solution to address carbon emissions and modernize vehicle controls. Batteries play a critical in the journey of electrification and modernization with battery voltage prediction as the foundation for safe and efficient operation. Due to its strong dependency on the prior information, the battery voltage was estimated with recurrent neural network methods in the recent literatures exploring variety deep learning techniques to estimate battery behaviors. In these studies, standard recurrent neural networks, gated recurrent units, and long-short term memory are popular neural network architectures under review. However, in most cases, each neural network architecture is individually assessed and therefore the knowledge about comparative study among three neural network architecture is limited. This paper presents a comparative study on the battery voltage predication using all three neural network architectures. In this study, all neural network

Cho, Gyouho, ZHU, DI, Campbell, Jeffrey

Three-way catalyst model development using physical and machine learning methods for engine control design purposes

2021-01-122109/21/2021

Gasoline engine control strategies ensure a combustion control around stoichiometry. That is because the three-way catalytic converter allows CO and HC oxidation under lean operating conditions while ensuring NOx reduction for rich mixtures. In case of engine malfunction, the controller must adapt to compensate for potential torque losse and other critical attributes, potentially leading to significant deviation of the fuel-air mixture richness from stoichiometry and larger emission levels. Therefore, during development of the engine fault diagnostics the impact on the pollutant emissions must be considered. In this paper, a model-based development process is proposed. It is based on system simulation modelling techniques, where a complete exhaust line is represented in order to predict tail-pipe emissions under stoichiometric, lean and rich conditions, for engine control design purposes. Two different modelling approaches are applied and evaluated in this paper. First, a physics-based

LOUSSAIEF, Sana, Akhmetov, Yerlan, GROISIL, Melanie, Takahashi, Daisuke, Kuhara, Masaki, Isobe, Yoshiki, Mas, Peter

Human Body Orientation from 2D Images

2021-01-008204/06/2021

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. One of the possible applications for this method is for social robots where the robot has

Abughalieh, Karam, Alawneh, Shadi

DA-IVE: MLP Based Data Association Method for Instantaneous Velocity Estimation Using Multi-Radar: An Experimental Validation Study

2021-01-009204/06/2021

This paper describes a novel Multi-Layer Perceptrons (MLP) learning-based association algorithm that is used in conjunction with an Instantaneous Velocity Estimator (IVE) to estimate the velocity of a surrounding vehicle using multi-radar sensors. The IVE algorithm requires at least two targets to be able to provide a velocity estimate. The approach suggested in this paper performs three stages of filtering on a list of targets available for the association to a given track. The algorithm identifies the one pair of targets that will provide the best instantaneous velocity estimation from all possible pairs. The three stages of filtering described ahead are, I - Semantic gating, II - MLP scoring, and III - Algebraic scoring. The IVE algorithm performs linear regression on the pair of targets it is finally provided to come up with a velocity estimation. This research also describes a novel method of labeling radar targets for use in the training of the neural network in association stage

Shakibajahromi, Bahareh, Krishnan, Anirudh Sarathy, Ati, Dilip, Jabalameli, Amirhossein, Kanzler, Steven, Shayestehmanesh, Saeed

Electronic Skin Sensor Decodes Human Motion

TBMG-3851702/01/2021

A powered, single-strained electronic skin sensor was developed that can capture human motion from a distance. The strain sensor, placed on the wrist, decodes complex five-finger motions in real time with a virtual 3D hand that mirrors the original motions. The deep neural network boosted by rapid situation learning (RSL) ensures stable operation regardless of its position on the surface of the skin.

Chemical Identification by Magneto-Elastic Sensing (ChIMES)

TBMG-3848302/01/2021

A low-cost sensor technology, called Chemical Identification by Magneto-Elastic Sensing (ChIMES), uses target response materials (TRMs) as actuators in magneto-elastic (M-E) sensors (Figure 1).

Earphone Tracks Facial Expressions, Even with a Facemask

TBMG-3818212/01/2020

Researchers have invented an earphone that can continuously track full facial expressions by observing the contour of the cheeks — and can then translate expressions into emojis or silent speech commands. With the ear-mounted device (called C-Face), users could express emotions to online collaborators without holding cameras in front of their faces — an especially useful communication tool as much of the world engages in remote work or learning.

Items per page:

1 – 50 of 391