Discrete Speed Research Articles

A Hierarchical Data-Partitioning Algorithm for Performance Optimization of Data-Parallel Applications on Heterogeneous Multi-Accelerator NUMA Nodes

Modern HPC platforms are highly heterogeneous with tight integration of multicore CPUs and accelerators (such as Graphics Processing Units, Intel Xeon Phis, or Field-Programmable Gate Arrays) empowering them to address the twin critical concerns of performance and energy efficiency. Due to this inherent characteristic, processing elements contend for shared on-chip resources such as Last Level Cache (LLC), interconnect, etc. and shared nodal resources such as DRAM, PCI-E links, etc., resulting in complexities such as resource contention, non-uniform memory access (NUMA), and accelerator-specific limitations such as limited main memory thereby necessitating support for efficient out-of-card execution. Due to these complexities, the performance profiles of data-parallel applications executing on these platforms are not smooth and deviate significantly from the shapes that allowed state-of-the-art load-balancing algorithms to find optimal solutions. In this paper, we propose a hierarchical two-level data partitioning algorithm minimizing the parallel execution time of data-parallel applications on clusters of $h$ identical nodes where each node has $c$ heterogeneous processors. This algorithm takes as input $c$ discrete speed functions of cardinality $m$ corresponding to the $c$ heterogeneous processors. It does not make any assumptions about the shapes of these functions. Unlike load balancing algorithms, optimal solutions found by the algorithm may not load-balance an application in terms of execution time. The proposed algorithm has low time complexity of $O(m^{2} \times h + m^{3} \times c^{3})$ unlike the state-of-the-art algorithm solving the same problem with the complexity of $O(m^{3} \times c^{3} \times h^{3})$ . We also propose an extension of the algorithm for clusters of $h$ non-identical nodes where each node has $c$ heterogeneous processors. We experimentally demonstrate the optimality of our algorithm using two well-known and highly optimized multi-threaded data-parallel applications, matrix-matrix multiplication and 2D fast Fourier transform, on a heterogeneous multi-accelerator NUMA node containing an Intel multicore Haswell CPU, an Nvidia K40c GPU, and an Intel Xeon Phi co-processor and a simulated homogeneous cluster of such nodes.

Geodynamics of very high speed transport systems

This work reveals the existence of a new dynamic load amplification mechanism due to ground surface loads. It is caused by the interaction between a moving vehicle's axle configuration and the vibration characteristics of the underlying soil-guideway system. It is more dominant than the traditionally considered ‘critical velocity’ dynamic amplification mechanism of the guideway-ground structure, and is of relevance to very high speed transport systems such as high speed rail.To demonstrate the new amplification mechanism, first a numerical model is developed, capable of simulating ground-wave propagation in the presence of a series of discrete high speed loads moving on a soil-guideway system. The model couples analytical equations for the transportation system guideway with the thin-layer element method for ground simulation. As a practical example, it is validated using high speed railroad field data and then used to analyse the response of a generalised single moving load at high speed. Next the effect of multiple discrete vehicle-guideway contact points is studied and it is shown that dynamic amplification is highly sensitive to load spacing when the speed is greater than the critical velocity. In particular, large resonant effects occur when the axle/magnet loading frequency and the propagating wave vibration frequency of the soil-guideway structure are equivalent. As an example, it is shown that for an individual case, although critical velocity might increase displacements by 50–100%, for the same scenario, axle configuration can increase displacements by 400%. It is also shown that resonance is sensitive to the total number of loading points and the individual frequencies excited by various spacings. The findings are important for current (e.g. high speed railway) and potential future (e.g. hyperloop) transport systems required to operate at speeds either close-to, or greater than the critical velocity of their supporting guideway-soil structure. In such situations, it is important to design the vehicle and supporting structure(s) as a combined system, rather than in isolation.

Turn‐to‐turn short circuit of motor stator fault diagnosis in continuous state based on deep auto‐encoder

In this study, a turn-to-turn short circuit of motor stator fault diagnosis system based on deep auto-encoder and soft-max classifier is proposed. It is also considered in the proposed system that an unbalanced power supply has similar current characteristics and is possible to be misidentified. The determination of neural network parameters and their effect in training are given systematically, which improves the performance compared with traditional auto-encoder. The proposed fault diagnosis system can map the preprocessed three-phase motor current signals to a two-dimensional vector, corresponding to certain area of the feature plane to identify different fault types. The proposed system is verified via simulation experiment using MATLAB and physical experiment using a motor in laboratory. The conclusion illustrates that when signals originating from discrete motor speed, load, and fault severity are used in training, the fault diagnosis is still effective in continuous state. The accuracy is 89% for speed verification and 95% for load verification in simulation and is above 99% in the physical experiment.

Discretely variable speed ratio control strategy for continuously variable transmission system considering hydraulic energy loss

By using the continuously variable transmission (CVT) to change the speed ratio continuously, the power resource components can operate in the high efficiency region for plug-in hybrid electric vehicle (PHEV) equipped with CVT, which improves the driving efficiency of powertrain. However, the frequent change of CVT speed ratio causes large energy loss of CVT hydraulic system and reduces the energy economy of PHEV. In view of this issue, the characteristic of energy loss for CVT hydraulic system is studied and the influence of speed ratio rate of change on energy loss of hydraulic system is obtained. Then, the simulation based on the continuously variable speed ratio control strategy (CVSRCS) is carried out and the results indicate that there is large energy loss of the CVT hydraulic system due to the frequent change of CVT speed ratio, which influences energy economy of the vehicle. Moreover, excessive driving jerk is generated, which significantly affects ride comfort of the vehicle. In order to reduce the adverse impact of frequent changes of CVT speed ratio on energy economy and ride comfort, a discretely variable speed ratio control strategy (DVSRCS) is proposed and discrete speed ratio is optimized by genetic algorithm. A comparative simulation for PHEV’s performance by adopting the CVSRCS and the proposed discretely variable speed ratio control strategy is carried out under a comprehensive driving cycle. The results of this study demonstrate that the proposed control strategy can not only significantly reduce the energy loss of CVT hydraulic system and enhance the energy economy, but also improve ride comfort.

Discrete Train Speed Profile Optimization for Urban Rail Transit: A Data-Driven Model and Integrated Algorithms Based on Machine Learning

Energy-efficient train speed profile optimization problem in urban rail transit systems has attracted much attention in recent years because of the requirement of reducing operation cost and protecting the environment. Traditional methods on this problem mainly focused on formulating kinematical equations to derive the speed profile and calculate the energy consumption, which caused the possible errors due to some assumptions used in the empirical equations. To fill this gap, according to the actual speed and energy data collected from the real-world urban rail system, this paper proposes a data-driven model and integrated heuristic algorithm based on machine learning to determine the optimal speed profile with minimum energy consumption. Firstly, a data-driven optimization model (DDOM) is proposed to describe the relationship between energy consumption and discrete speed profile processed from actual data. Then, two typical machine learning algorithms, random forest regression (RFR) algorithm and support vector machine regression (SVR) algorithm, are used to identify the importance degree of velocity in the different positions of profile and calculate the traction energy consumption. Results show that the calculation average error is less than 0.1 kwh, and the energy consumption can be reduced by about 2.84% in a case study of Beijing Changping Line.

Predictive Maneuver Planning for an Autonomous Vehicle in Public Highway Traffic

This paper outlines a predictive maneuver-planning method for autonomous vehicle navigating public highway traffic. The method integrates discrete maneuvering decisions, i.e., lane and reference speed selection automata, with a model predictive control-based motion trajectory-planning scheme. A key notion is to apply a predictive reference speed pre-planning for each lane at each time step of a selected prediction horizon. This is done based on the predicted likely motion of the autonomous vehicle and other object vehicles subject to sensor noise and environmental disturbances. Then, an optimization problem is configured that computes safe, sub-optimal plans for the trajectories of both the motion states (and inputs) and maneuver references for the prediction horizon to accomplish maneuvers like lane keeping, lane change, or obstacle avoidance. While a first formulation of the problem results in a mixed-integer nonlinear programming problem, it is shown that a relaxation can be adopted that reduces the computational complexity to a low-order polynomial time nonlinear program that can be solved more efficiently. Through simulation of a series of multi-lane highway scenarios and comparison with one-maneuver planning approach and an adaptive cruise control approach, the proposed predictive maneuver planning is illustrated to better accommodate the traffic environment with feasible execution time. Also, the reference speed pre-planning improves the optimality and the robustness of the maneuver decision in trajectory planning without adding computational complexity to the optimization problem.

Optimization of tow-steered composite wind turbine blades for static aeroelastic performance

The concept of passive aeroelastic tailoring is explored to maximize power extraction from an NREL 5-MW wind turbine blade rotating in a uniform flow, where a single blade with periodic boundary conditions simulates a three-bladed wind turbine rotor. Variable-angle tow composite materials model the spanwise-variable wind turbine blade design and enable coupled bend-twist deformations under aerodynamic loading. A constrained optimization algorithm varies only the composite fiber angles along the blade span to determine their optimal distribution at four uniform inflow conditions ranging from cut-in to rated wind speeds. The computational fluid dynamics solver CRUNCH CFD ® and commercial finite element analysis solver Abaqus compute the static aerodynamic loads and structural deformations of the blade, respectively, until aeroelastic convergence is achieved. The resulting computational aeroelastic formulation predicts an increase in turbine power extraction by up to 14% when the blade is optimized near the cut-in wind speed, and by 7% when optimized at the rated wind speed. Using the results from optimizations at discrete wind speeds, two blade design strategies are evaluated to determine a single composite layup for the blade that maximizes power output over the range of operational wind speeds.

Elucidating how correlated operation of shear transformation zones leads to shear localization and fracture in metallic glasses: Tensile tests on Cu[sbnd]Zr based metallic-glass microwires, molecular dynamics simulations, and modelling

The plastic deformation of metallic glasses (MG) is well-known to occur via shear transformation zones (STZs) on the scale of atomic clusters, yet fracture of MG takes place via shear bands of the micron scale. So far, understanding on how the operation of STZs leads to shear localization and fracture remains limited. In this work, tensile tests on Cu/Zr-based MG micro-wires show that both the first-yield and fracture stress exhibit the Weibull distribution, and fractography reveals that shear localization in the form of intense shear bands leads to shear fracture. Molecular dynamics (MD) simulations show that shear bands form via the correlated emergence and operation of discrete STZs close to one another. To describe how the stochastic yet correlated occurrence of the discrete STZs leads to shear localization, a model is constructed to relate the probability of the successive operation of discrete STZs, to their nucleation density. The model predicts that, if nucleation density of the STZs grows along the strain path, as prior shear events triggers the emergence of new STZs, then successive occurrence of discrete shear events speeds up rapidly to an asymptotic state which is exactly the condition of shear localization and shear banding. Furthermore, the MD results suggest an exponential growth law for the occurrence of the STZs along the strain path, which also gives predictions in good agreement with the experimental Weibull distributions of the first-yield stress of Cu/Zr-based MGs.

Kinematic Correlates of Kinetic Outcomes Associated With Running-Related Injury.

High magnitudes and rates of loading have been implicated in the etiology of running-related injuries. Knowledge of kinematic variables that are predictive of kinetic outcomes could inform clinic-based gait retraining programs. Healthy novice female runners ran on a treadmill while 3-dimensional biomechanical data were collected. Kinetic outcomes consisted of vertical impact transient, average vertical loading rate, instantaneous vertical loading rate, and peak braking force. Kinematic outcomes included step length), hip flexion angle at initial contact, horizontal distance from heel to center of mass at initial contact, shank angle at initial contact, and foot strike angle. Stepwise multiple linear regression was used to evaluate the amount of variance in kinetic outcomes explained by kinematic outcomes. A moderate amount of variance in kinetic outcomes (vertical impact transient = 46%, average vertical loading rate = 37%, instantaneous vertical loading rate = 49%, peak braking force = 54%) was explained by several discrete kinematic variables-predominantly speed, horizontal distance from heel to center of mass, foot strike angle, and step length. Hip flexion angle and shank angle did not contribute to any models. Decreasing step length and transitioning from a rearfoot strike may reduce kinetic risk factors for running-related injuries. In contrast, clinical strategies such as modifying shank angle and hip flexion angle would not appear to contribute significantly to the variance of kinetic outcomes after accounting for other variables.

The bullet problem with discrete speeds

Bullets are fired from the origin of the positive real line, one per second, with independent speeds sampled uniformly from a discrete set. Collisions result in mutual annihilation. We show that a bullet with the second largest speed survives with positive probability, while a bullet with the smallest speed does not. This also holds for exponential spacings between firing times. Our results imply that the middle-velocity particle survives with positive probability in a two-sided version of the bullet process with three speeds known to physicists as ballistic annihilation.

Travel Time Functions Prediction for Time-Dependent Networks

The studies on the TDN (time-dependent network), in which the travel time of the same road segment varies depending on the time of the day, have attracted much attention of researchers, but there is little work focusing on the travel time functions prediction problem. Though traditional methods for travel time or travel speed prediction problem can be used to generate the travel time functions, they have some limitations due to the need of less breakpoints, fine granularity, and long-term prediction. In this paper, we study the travel time functions prediction problem for TDN based on taxi trajectory data. In order to maintain a high degree of accuracy in fine-grained and long-predicted situations, we take into account not only the traffic incidents but also the data sparsity. Specifically, a traffic incident detection method is proposed based on k-means algorithm and a downstream-based strategy is proposed to estimate the speeds of segments considering the data sparsity. To make the breakpoints of function not so much, a prediction algorithm based on classification using ELM (extreme learning machine) is proposed, which predicts the speed classes taking both the weather and the adjacent segment conditions into account. In addition, a transformation method is presented to convert the discrete travel speeds into piecewise linear functions satisfying FIFO (First-In-First-Out) property. The experimental results show that ELM outperforms SVM (support vector machine) with regard to both the training time and prediction accuracy. Moreover, it also can be seen that both the weather conditions and the adjacent segment conditions have impact on the prediction accuracy.

A Novel Data-Partitioning Algorithm for Performance Optimization of Data-Parallel Applications on Heterogeneous HPC Platforms

Modern HPC platforms have become highly heterogeneous owing to tight integration of multicore CPUs and accelerators (such as Graphics Processing Units, Intel Xeon Phis, or Field-Programmable Gate Arrays) empowering them to maximize the dominant objectives of performance and energy efficiency. Due to this inherent characteristic, processing elements contend for shared on-chip resources such as Last Level Cache (LLC), interconnect, etc. and shared nodal resources such as DRAM, PCI-E links, etc. This has resulted in severe resource contention and Non-Uniform Memory Access (NUMA) that have posed serious challenges to model and algorithm developers. Moreover, the accelerators feature limited main memory compared to the multicore CPU host and are connected to it via limited bandwidth PCI-E links thereby requiring support for efficient out-of-card execution. To summarize, the complexities (resource contention, NUMA, accelerator-specific limitations, etc.) have introduced new challenges to optimization of data-parallel applications on these platforms for performance. Due to these complexities, the performance profiles of data-parallel applications executing on these platforms are not smooth and deviate significantly from the shapes that allowed state-of-the-art load-balancing algorithms to find optimal solutions. In this paper, we formulate the problem of optimization of data-parallel applications on modern heterogeneous HPC platforms for performance. We then propose a new model-based data partitioning algorithm, which minimizes the execution time of computations in the parallel execution of the application. This algorithm takes as input a set of $p$ discrete speed functions corresponding to $p$ available heterogeneous processors. It does not make any assumptions about the shapes of these functions. We prove the correctness of the algorithm and its complexity of $O(m^3 \times p^3)$ , where $m$ is the cardinality of the input discrete speed functions. We experimentally demonstrate the optimality and efficiency of our algorithm using two data-parallel applications, matrix multiplication and fast Fourier transform, on a heterogeneous cluster of nodes where each node contains an Intel multicore Haswell CPU, an Nvidia K40c GPU, and an Intel Xeon Phi co-processor.

Optimal task execution speed setting and lower bound for delay and energy minimization

The current technology trend reveals that static power consumption is growing at a faster rate than dynamic power consumption. In this paper, energy-efficient task scheduling is studied when static power consumption is a significant part of energy consumption which cannot be ignored. The problems of scheduling a set of independent sequential tasks on identical processors so that the schedule length is minimized for a given energy consumption constraint or the energy consumption is minimized for a given schedule length constraint are investigated. For a given schedule, the optimal task execution speed setting for delay and energy minimization is found analytically. Lower bounds for the minimum schedule length of a set of tasks with a given energy consumption constraint and the minimum energy consumption of a set of tasks with a given schedule length constraint are established. Our lower bounds are applicable to sequential or parallel, and independent or precedence constrained tasks, on processors with discrete or continuous speed levels, and bounded or unbounded speed ranges. The significance of these lower bounds is that they can be used to evaluate the performance of any heuristic algorithms when compared with optimal algorithms. Experimental study on the performance of list scheduling algorithms is performed and it is shown that their performance is very close to the optimal. To the best of the author’s knowledge, this is the first paper that provides such analytical results for energy-efficient task scheduling with both dynamic and static power consumptions.

Improvement of jaw crushers reliability using elastic pneumatic elements in the connection of kinematic pairs

Downtimes associated with the replacement of the plain bearing liners because of their tear and wear are among the numerous technical reasons for relatively short-term but rather frequent downtimes in the conditions of operation of jaw crushers. The rapid wear of the liners is the result not only of contact friction between axle journals and liners, but also the action of dynamic forces. Dynamic forces that arise during the operation of jaw crushers are due to the presence of a gap in the kinematic pairing of the links (journals and bearing lining) and discrete speed values of the relative movement of the links inside the clearances of plain bearings. Thus, the reliable operation of the machine largely depends on the creation of conditions that ensure a gapless contact of the conjugated links. It was experimentally proved that the use of mechanisms for the selection of gaps with elastic elements, which during the whole cycle of operation of the crank and rocker mechanism select a gap in the joints of kinematic pairs, prevent the appearance of additional dynamic loads and, thereby, increase the reliability of the jaw crushers.

Hybrid Computational Mechanical Sensorless Fuzzified Technique for Speed Estimation of Permanent Magnet Direct Current Brushed Motor

This paper proposes a novel fuzzified mechanical sensorless speed estimation technique for permanent magnet direct current (PMDC) brushed motor operated with pulse width modulation (PWM) based terminal voltage. For replacement of conventional mechanical speed sensors, few mechanical sensorless speed estimation techniques have been reported recently in the field of PMDC brushed motor. In this paper, a novel technique is applied over a recently reported semianalytical dynamic time-domain PMDC brushed motor model incorporating space-domain effects, namely slotting effect and commutation phenomenon to simulate the proposed estimation process. Ripple and back electromotive force based standalone estimation approaches are hybridized in order to eliminate uncertainties in all regions of estimation. In this paper, a discrete fuzzified hybrid speed estimation model has been proposed with adequate dynamic simulation responses. Further the proposed discrete speed estimation model has been applied experimentally over a PWM-driven 24-V, 12-teeth, 100-W PMDC brushed motor and various dynamic real-time responses have been compared with conventionally captured speed samples to validate the proposed model.

Application of finitized power series distributions to accelerated variate generation. Part II: the case of the logarithmic distribution

AbstractThe method of aliasing vastly expands the palette of discrete random variate generation methodologies while providing excellent speed. However, its application is limited to finitely supported distributions. We demonstrate that an application of moment preserving finitization called the Negative Taylor Series Finitization (NTSF) method for the power series family of discrete distributions, when coupled with the method of aliasing, can greatly improve infinitely supported discrete random variate generation speed with certain limitations. We illustrate this with the logarithmic power series distribution, and we compare four published algorithms designed to generate random variates from a logarithmic distribution to the aliasing method of random variate generation from an NTSF version of the same distribution. We compare the accuracy and speed (user‐time) of these various methods for generating variates from a logarithmic distribution.

Free Convection Heat Transfer of Nanofluids into Cubical Enclosures with a Bottom Heat Source: Lattice Boltzmann Application

The main purpose of this work is to investigate the hydrodynamic and thermal characteristics of an Ag-water nanofluid within a cubical enclosure, including a heat source which located at the center of the bottom wall. Due to its crucial role in the characterization of the main transfer into such configurations, the impact of some parameters is widely inspected. It consists the Rayleigh number (103 to 106), the nanoparticles volume fraction (0% to 10%), the width (10% ≤ w ≤ 40%) and the height of the heat source (10% ≤ h ≤ 50%). To do so, a numerical code based on the Lattice-Boltzmann method, coupled with a finite difference one, is utilised. The latter has been validated after comparison between the present results and those of the literature. It is to note, at the end, that the three dimensions D3Q19 model is adopted based on a cubic Lattice, where each pattern of the latter is characterized by nineteen discrete speeds.

Discrete Adaptive Speed Sensorless Drive of Induction Motors

The use of mechanical speed sensor in the field of control of induction motors can cause some problems. Recently, many software are developed to estimate the speed by using only electrical measurements and machine parameters. This paper presents a new speed sensorless drive of induction motors for field oriented control. The proposed speed estimator is based on the Model Reference Adaptive Scheme (MRAS) and uses based the comparison of estimated and measured stator current. The reduction error is achieved by an appropriate unity transfer function as reference model. The estimator formulation is defined and developed in the discrete case. The proposed algorithm requires neither matrix calculation nor approximation on Taylor series of matrix exponential. To establish the stability of the discrete estimator, the hyperstability concept introduced by Popov in the discrete case is used. The influence of the sampling period on the algorithm stability of is studied. Some simulation results are presented, in order to valid the theoretical development.

Flow shop for dual CPUs in dynamic voltage scaling

We study the following flow shop scheduling problem on two processors. We are given n jobs with a common deadline D, where each job j has workload pi,j on processor i and a set of processors which can vary their speed dynamically. Job j can be executed on the second processor if the execution of job j is completed on the first processor. Our objective is to find a feasible schedule such that all jobs are completed by the common deadline D with minimized energy consumption. For this model, we present a linear program for the discrete speed case, where the processor can only run at specific speeds in S={s1,s2,⋯,sq} and the job execution order is fixed. We also provide a mα−1-approximation algorithm for the arbitrary order case and for continuous speed model where m is the number of processors and α is a parameter of the processor.We then introduce a new variant of flow shop scheduling problem called sense-and-aggregate model motivated by data aggregation in wireless sensor networks where the base station needs to receive data from sensors and then compute a single aggregate result. In this model, the first processor will receive unit size data from sensors and the second processor is responsible for calculating the aggregate result. The second processor can decide when to aggregate and the workload that needs to be done to aggregate x data will be f(x) and another unit size data will be generated as the result of the partial aggregation which will then be used in the next round aggregation. Our objective is to find a schedule such that all data are received and aggregated by the deadline with minimum energy consumption. We present an O(n5) dynamic programming algorithm when f(x)=x and a greedy algorithm when f(x)=x−1.Finally, we investigate the performance of the flowshop problem when the order of jobs is fixed by comparing it to the approximation algorithm with an arbitrary order. We show experimentally that the approximation ratio is close to 1 when there are few machines and when there are more jobs.

Energy-efficient real-time scheduling for two-type heterogeneous multiprocessors

We propose three novel mathematical optimization formulations that solve the same two-type heterogeneous multiprocessor scheduling problem for a real-time taskset with hard constraints. Our formulations are based on a global scheduling scheme and a fluid model. The first formulation is a mixed-integer nonlinear program, since the scheduling problem is intuitively considered as an assignment problem. However, by changing the scheduling problem to first determine a task workload partition and then to find the execution order of all tasks, the computation time can be significantly reduced. Specifically, the workload partitioning problem can be formulated as a continuous nonlinear program for a system with continuous operating frequency, and as a continuous linear program for a practical system with a discrete speed level set. The latter problem can therefore be solved by an interior point method to any accuracy in polynomial time. The task ordering problem can be solved by an algorithm with a complexity that is linear in the total number of tasks. The work is evaluated against existing global energy/feasibility optimal workload allocation formulations. The results illustrate that our algorithms are both feasibility optimal and energy optimal for both implicit and constrained deadline tasksets. Specifically, our algorithm can achieve up to 40% energy saving for some simulated tasksets with constrained deadlines. The benefit of our formulation compared with existing work is that our algorithms can solve a more general class of scheduling problems due to incorporating a scheduling dynamic model in the formulations and allowing for a time-varying speed profile.