Accuracy and Efficiency Assessment of Harmonic Balance Method for Unsteady Flow in Multi-Stage Turbomachinery

With the relative movement of neighboring blade rows,flows in multi-stage turbomachinery are unsteady and periodic in time at the design condition.As an alternative to the widely used time domain time marching method,the harmonic balance(HB)method has been successfully applied to simulate the essentially unsteady flow of multi-stage turbomachinery.By modelling various number of harmonics,the accuracy of this method could be adjusted at different level of computational cost.In this article,accuracy of the harmonic balance method is not only validated against the time domain time marching method,as in most previous works on this topic,but also against the data from an experiment campaign of a two-stage high-pressure turbine where strong tip leakage flow exists.Efficiency of this method is also assessed in detail by adjusting the number of harmonics and comparing with time domain time marching solution results.Results show that the harmonic balance method is a flexible tool with adjustable accuracy for fast-turnaround unsteady flow simulation of multi-stage turbomachinery.Results from this work can provide a guidance in applying the harmonic balance method with balance between accuracy and computational cost.

Impact of Motor Stator Winding Faults on Motor Differential-mode Impedance and Mode Transformation

Motor impedance and mode transformation have significant effects on the electromagnetic interference(EMI)generated in motor drive systems.Stator winding faults commonly cause motor failure;however,in their early stages,they may not affect the short-term operation of the motor.To date,EMI noise under the influence of premature stator winding faults has not been adequately studied,particularly the differential-mode(DM)noise due to the common-mode(CM)-to-DM transformation.This study investigates and quantifies the influence of stator winding faults on the motor DM impedance and mode transformation.First,the transmission line model of an induction motor is described based on the scattering(S)parameter measurements of each phase of the motor.It offers the flexibility to emulate different types of stator winding faults at specific locations and various severities,such that the impacts of the faults on the motor DM impedance can be easily estimated.Second,a test setup is proposed to quantify the CM-to-DM transformation due to the stator winding faults.The findings of this study reveal that even the early stages of stator winding faults can result in significant changes in the DM noise.

机器学习下的基于退化特征提取和时间平滑分析的非平稳工业过程在线故障预测

故障退化预测是预估过程劣化和故障发生的时间,已被认为是维护策略中的一个关键组成部分。故障退化过程通常是缓慢变化的,可以用自回归模型来建模。然而,工业过程往往表现出典型的非平稳特性,这就给故障退化信息的获取和非平稳过程的建模带来了挑战。针对上述问题,本文提出了一种新的故障退化建模和在线故障预测策略。首先,提出一种面向故障退化的慢特征分析(FDSFA)算法提取故障退化方向,并沿该方向提取候选故障退化特征。然后,利用趋势评估算法来选择主要的故障退化特征。其次,结合主要的故障退化特征及其稳定性加权因子,计算关键故障退化因子来表征故障退化趋势。针对过程非平稳特性,建立了带时序平滑正则项的时变回归模型。在更新策略的基础上,通过对预测误差的分析和建模,进一步建立了在线故障预测模型。最后,通过一个实际的工业案例验证了所提方法的预测性能。

A Multi-Scale Thermal Analysis Method for Data Centers with Application in a Ship Data Center

Thermal analysis of data centers is in urgent need to ensure that computer chips remain below the critical temperature while the energy consumption for cooling can be reduced.It is difficult to obtain detailed hotspot locations and temperatures of chips in large data centers containing hundreds of racks or more by direct measurement.In this paper,a multi-scale thermal analysis method is proposed that can predict the temperature distribution of chips and solder balls in data centers.The multi-scale model is divided into six scales:room,rack,server,Insulated-Gate Bipolar Transistor(IGBT),chip and solder ball.A concept of sub-model is proposed and the six levels are organized into four simulation sub-models.Sub-model 1 contains Room,Rack and Server(RRS);Sub-model 2 contains Server and IGBT(SI);Sub-model 3 contains IGBT and Chip(IC),and Sub-model 4 contains Chip and Solder-ball(CS).These four sub-models are one-way coupled by transmitting their results as boundary conditions between levels.The full-field simulation method is employed to verify the efficiency and accuracy of multi-scale simulation method for a single-rack data center.The two simulation results show that the highest temperature emerges in the same location.The Single-rack Full-field Model(SRFFM)costs 2.5 times more computational time than that with Single-rack Multi-scale Model(SRMSM).The deviation of the highest temperature of chips and solder balls are 1.57℃and 0.2℃between the two models which indicates that the multi-scale simulation method has good prospect in the data center thermal simulation.Finally,the multi-scale thermal analysis method is applied to a ship data center with 15 racks.

具有潜在标签一致性和稀疏高斯特征学习的非线性工业过程故障诊断方法

随着工业过程复杂程度的不断提高,高维工业数据呈现出很强的非线性,给工业过程的故障诊断带来了相当大的挑战。为了有效提取对故障诊断至关重要的深层特征,首先,设计了稀疏高斯特征提取器(SGFE)来学习非线性映射,该映射能够将原始数据投影到具有故障标签维度的特征空间中。特征空间由故障类别标签的one-hot编码作为正交基构成。这样,SGFE可以从原始数据中逐步学习到与故障标签相关的深度稀疏高斯特征。然后,在特征空间中,设计了稀疏高斯(SG)损失函数,将特征约束为多个稀疏多元高斯分布。稀疏的高斯特征在特征空间是线性可分的,这有利于提高下游故障分类任务的性能。最后,对本文所提出的SGFE算法的可行性和实用性分别通过手写数字(MNIST)数据集和田纳西-伊士曼(TE)过程进行了验证。

Adjoint-Based Geometrically Constrained Aerodynamic Optimization of a Transonic Compressor Stage

Efficient method to handle the geometric constraints in the optimization of turbomachinery blade profile is required. Without constraints on the blade thickness, optimal designs typically yield thinner blade to reduce the friction loss, however, at the risk of degraded strength and stiffness. This issue is seldom discussed and existing literature always treat the blade thickness constraint in an indirect manner. In this work, two different geometric constraints on the blade thickness are proposed and applied in the adjoint optimization: one is on the maximum blade thickness and the other is on the blade area. Methods to compute sensitivities of both constraints are proposed and they are integrated into an optimization system based on a finite volume code and a solver for the discrete adjoint equation. Adjoint optimization is conducted to minimize the entropy production in a transonic compressor stage. Results for the adjoint optimization without geometry constraint and two comparative cases are detailed. It is indicated that three cases yield similar performance improvement;however, if geometry constraints are properly handled, the optimal designs have almost the same maximum thickness as the original design, compared to a thinner blade profile with 14% reduction of maximum thickness in the case without geometry constraint. The cases considering geometry constraints also consume slightly reduced Central Processing Unit(CPU) cost. Result of this work verifies the effectiveness of the proposed method to treat geometric constraints in adjoint optimization.

Effect of Unsteady Secondary Vortices on Performance of a 1.5-Stage High-Pressure Turbine

In modern high-load high-pressure turbine,the secondary flow in the blade channel is very strong and occupies a large spanwise region.Although high-quality experimental data at the stage interfaces have been obtained in previous research,the influence of the clocking position on the secondary flow patterns is not fully understood.This paper investigates the clocking effect in a 1.5-stage high-pressure turbine and focuses on the variations of secondary flow patterns and their effect on the turbine performance.The detailed flow fields of various clocking positions were obtained by carrying out unsteady flow simulations using an in-house code.Among the four clocking positions in this work,the highest entropy generation was observed when the wakes from stator 1 hit the leading edges of stator 2,which is opposite to the well-known conclusion for the turbine with high-aspect-ratio blades.Detailed flow analysis showed that the wakes and the near tip secondary vortices from stator 1 showed different traces when entering the stator 2 channel and the secondary vortices clearly have a more important influence in determining the performance.The different behaviors of the secondary vortices explained the performance variations due to the clocking effect.

Physics Informed Neural Network-based High-frequency Modeling of Induction Motors

The high-frequency(HF)modeling of induction motors plays a key role in predicting the motor terminal overvoltage and conducted emissions in a motor drive system.In this study,a physics informed neural network-based HF modeling method,which has the merits of high accuracy,good versatility,and simple parameterization,is proposed.The proposed model of the induction motor consists of a three-phase equivalent circuit with eighteen circuit elements per phase to ensure model accuracy.The per phase circuit structure is symmetric concerning its phase-start and phase-end points.This symmetry enables the proposed model to be applicable for both star-and delta-connected induction motors without having to recalculate the circuit element values when changing the motor connection from star to delta and vice versa.Motor physics knowledge,namely per-phase impedances,are used in the artificial neural network to obtain the values of the circuit elements.The parameterization can be easily implemented within a few minutes using a common personal computer(PC).Case studies verify the effectiveness of the proposed HF modeling method.