Evaluation and Compensation of Quantization Effect on Servo Performance for Hard Disk Drives

We investigated the quantization effect on the servo performance of a 1.8-in hard disk drive by changing the bits of the proposed quantizer model. We measured and analyzed the frequency response of the quantizer with different bits. The corresponding error rejection functions show that the poor rejection ability of the servo loop to low-frequency disturbances is caused by the quantizer in addition to the actuator pivot friction behavior. We propose a simple and low-cost scaling scheme to compensate the effect of the quantizer. With the compensation, the effect on the rejection ability is mostly due to the friction. Therefore, the effects from quantization and friction on the error rejection function can be differentiated, and the quantization and friction induced problems can be dealt with separately in the servo loop. In addition, through the proposed quantization model and measurement methodology, suitable bit resolution for the quantizer with and without the compensation can be identified.     

Noninvasive Vibration Diagnostics to Detect Head–Disk Contact in Hard Disk Drives

Reliability of modern hard disk drives (HDDs) is a major concern especially with decreasing head-to-disk spacing in order to attain higher recording densities. One of the reasons for HDD failure is the occurrence of intermittent or sustained head–disk contacts. It is therefore important to be able to detect such head–disk contacts early on so that backup safety schemes can be implemented before catastrophic HDD failure and permanent data loss. In the study reported here, we used noninvasive vibration diagnostics and multiaxial vibration measurements to assess the condition and functionality of HDDs. We chose a miniature triaxial accelerometer for the diagnostic study, given that it is relatively inexpensive and easy to use. We measured contact vibration, induced by a scratch on the top surface of the rotating disk, on the outside of the HDD by attaching an accelerometer on the actuator shaft. We found that contact vibrations between the head and disk inside the HDD can reliably be measured on the outside of the HDD with an accelerometer. We also found that a miniature accelerometer does not cause significant mass loading, as verified by noncontacting laser vibrometry.     

Evaluating Track-Following Servo Performance of High-Density Hard Disk Drives Using Patterned Media

In recent years, track densities of magnetic hard disks have continued to grow. A promising approach to continuing the trend to ultrahigh density is using bit-patterned media (BPM). However, the implementation of BPM in hard disk drives (HDDs) to achieve high recording density is challenging and requires various new techniques, such as new servo pattern designs and position error signal (PES) decoding schemes. In applying BPM in HDDs, it is important to select a servo pattern providing sufficient PES quality for head positioning. In this paper, we discuss evaluation of PES quality and servo pattern performance from a closed-loop (servo) point of view in order to evaluate the quality of several servo patterns. We consider three servo patterns (the amplitude pattern, chevron pattern, and differential frequency pattern) as case studies. We developed a PES simulation tool to provide a realistic HDD track-following simulation. Because of PES nonlinearity in the amplitude servo pattern, we considered time-based servo patterns as alternatives. For time-based servo patterns, we found that readback signal sampling and transition jitter greatly affect PES quality. Therefore, we conclude that the differential frequency servo pattern is superior to other patterns, since it is less sensitive to transition jitter and readback signal sampling.      

Failure Prediction,Lead Time Estimation and Health Degree Assessment for Hard Disk Drives Using Voting Based Decision Trees

Hard Disk drives (HDDs) are an essential component of cloud computing and big data, responsible for storing humongous volumes of collected data. However, HDD failures pose a huge challenge to big data servers and cloud service providers. Every year, about 10% disk drives used in servers crash at least twice, lead to data loss, recovery cost and lower reliability. Recently, the researchers have used SMART parameters to develop various prediction techniques, however, these methods need to be improved for reliability and real-world usage due to the following factors: they lack the ability to consider the gradual change/deterioration of HDDs; they have failed to handle data unbalancing and biases problem; they don’t have adequate mechanisms for health status prediction of HDDs. This paper introduces a novel voting-based decision tree classifier to cater failure prediction, a balance splitting algorithm for the data unbalancing problem, an advanced procedure for lead time estimation and R-CNN based approach for health status estimation. Our system works robustly by considering a gradual change in SMART parameters. The system is rigorously tested on 3 datasets and it delivered benchmarks results as compared to the state of the art.     

磁悬浮硬盘转子控制系统中时序电路测试技术

为提高磁悬浮硬盘转子控制系统的可靠性，提出了对时序芯片的测试方法：手工测试和计算机辅助测试；及时时序电路逻辑功能的两种检测技术；基于状态表的测试技术和自动检测技术。对于时序芯片，手工测试方法效率较低，计算机辅助测试方法适用范围更广；对于电路逻辑功能，基于状态表的测试技术操作相对困难，但无须增加额外测试电路；自动检测技术操作简单，且测试速度快，但需要增加辅助检测电路。时序电路测试技术对保证悬浮硬盘转子各组成部分协同工作十分重要。     

Modified Damping Scheduling Proximate Time Optimal Servomechanism for Improvements in Short Strokes in Hard Disk Drives

Proximate time-optimal servomechanism (PTOS) has been widely used as a point-to-point motion control algorithm in many control systems, especially in hard disk drives, because of its simplicity and good performance. A great deal of research for improving PTOS has been reported; however, most of it is complex or needs a time-consuming tuning process. Damping scheduling PTOS (DSPTOS), which maintained the simplicity of PTOS and showed faster settling in long strokes by scheduling a closed-loop damping coefficient, was proposed previously. In this paper, we propose a modified DSPTOS which can schedule a closed-loop damping coefficient not only in long strokes, but also in short strokes. The proposed algorithm improves seek and settling performances significantly in short strokes. The effectiveness of the proposed algorithm is verified through simulations and real-time experiments. Consequently, we achieved over a 25% settling time improvement for a short stroke (10 tracks).     

基于自适应模糊神经网络的DMFC温度建模和控制

针对DMFC电堆的实时控制要求，应用自适应模糊神经网络技术对DMFC电堆的工作温度进行辨识建模和控制。在温度控制过程中，将训练好的网络模型作为DMFC电堆控制系统的参考模型，并对控制模型的参数进行在线自适应调整。仿真结果表明所设计的自适应模糊神经控制器性能优越。     

基于神经网络的智能诊断

人工智能与诊断理论的结合形成了智能诊断,早期发展的模拟人脑思维推理的、基于知识的专家系统以串行运行的格式进入设备诊断领域,形成了基于知识的诊断推理专家系统,国内外已有许多成熟的商品化软件系统。近几年新发展起来的人工智能的一个分支－－人工神经网络模仿人脑物理结构以其强大的并行运算和联想能力非常适合于设备诊断中状态识别,本单位研制的通用型神经网络智能诊断系统,已达到商品化水平,并已在生产线上运行。     

MCFC型燃料电池的神经网络建模与基于FGA的模糊控制研究

为了提高燃料电池的发电性能,熔融碳酸盐燃料电池(MCFC)堆的运行温度应该控制在一个合适的范围内。本文首先利用RBF神经网络辨识复杂非线性系统的能力,基于实验的输入输出数据,建立起MCFC电堆的神经网络温度模型;然后设计了MCFC电堆工作温度的一个基于模糊遗传算法的在线模糊控制器,用模糊遗传算法同时优化模糊控制器的参数及规则。最后用神经网络的辨识模型代替实际的电堆进行控制仿真,仿真结果证明建模是有效的,所设计的模糊控制器具有较好的性能。     

基于神经网络的组播路由调度方法

针对在高速包交换计算机网络中端到端时延及时延抖动限制的组播路由问题，提出了基于Ｈｏｐｆｉｅｌｄ神经网络的组播路由优化算法。实验表明，本算法能根据组播应用对时延及时延抖动限制的要求，快速，有效地构造最优组播树，有较强的实时性。     

基于深度循环神经网络的社交网络用户情感研究

目的 将深度学习与社交网络、情感计算相结合,探索利用深度神经网络进行社交网络用户情感研究的新方法和新技术,探索模型在用户需求分析和推荐上的应用.方法 自动筛选和挖掘海量社交网络数据,研究具有长时记忆的非先验情感预测方法,对网络中海量的用户数据、人与人之间关系进行建模,为关联时间序列创建LSTM模型,并结合其相互关系融入统一的大型深度循环网络中.具体包括:基于注意力模型的社交网络异构数据处理;基于深度LSTM的长时记忆建模,研究子网络选取、深度LSTM设计,以及针对社交网络的大型网络结构设计;基于社交网络情感模型和强化学习的推荐算法.结果 提高了分析的准确度,降低了对先验假设的依赖,减轻了人工情感模型的工作量和偏差,增强了对不同网络数据的普适性;供深度模型使用.结论 研究成果促进了深度学习与情感计算的结合,可推动网络用户行为分析和预测的研究,可用于个性化推荐、定向广告等领域,具有广泛的学术意义和应用前景.     

基于人工神经网络的结构损伤识别方法

根据观测的结构频率变化,建立了基于人工神经网络的悬臂梁结构损伤识别方法.把结构固有频率的变化率作为BP神经网络的输入参数,对悬臂梁结构模型进行了损伤数值模拟计算.为了提高神经网络的泛化推广能力和收敛速度,将BFGS优化方法应用到神经网络的训练过程中.数值计算结果表明,所建立的结构损伤识别方法具有收敛速度快、识别精度高等特性.     

基于切片谱和神经网络的旋转机械故障诊断方法

RBF神经网络是仿生物神经建立的人工模式识别模型,它具有良好的泛化能力和分类能力;切片谱是对双相干谱算法的简化,同样可以用来定量描述振动信号的非线性相位耦合特征.本文结合双谱和RBF神经网络,以信号的切片谱作为故障特征量,以RBF神经网络作为分类器实现旋转机械故障诊断;最后对轴承故障进行仿真实验,结果表明,本方法能有效对轴承故障模式进行分类.     

神经网络方法在自相关过程控制中的应用

将传统休哈特控制图应用于自相关过程控制时,会引发大量虚发报警.本文将使用时间序列模型模拟自相关过程并将神经网络方法引入自相关过程控制中.以神经网络特有的模式识别技术,对自相关过程中均值发生突变的情况进行监控,取得了良好效果.     

采用细胞神经网络快速校正生物芯片图像倾斜

倾斜校正是全自动生物芯片图像处理必不可少的环节。针对以往倾斜校正算法耗时过长,本文提出一种快速倾斜校正算法。算法首先用细胞神经网络（CNN）寻求各个样点的中心,然后进行Hough变换,再计算方差来寻求倾斜角,最后利用CNN灰度图像旋转模板进行倾斜校正。本算法利用了细胞神经网络并行处理的特性,并充分考虑了生物芯片图像的特点。理论分析与试验结果显示本文算法能够准确高速地完成倾斜校正。     

基于前馈预测的电液伺服PID控制系统

根据预测控制思想,针对已有的电液伺服调速系统模型,设计了一种基于前馈预测的PID控制系统。该控制系统不依赖对象的精确数学模型,利用预测控制的快速性以及自适应性等特点,通过滚动优化与PID相结合来确定系统当前时刻的控制量。仿真实验结果表明,该文建立的新型控制器能够满足现有的电液伺服调速系统的控制性能要求,实现对该系统的快速精确控制。     

Nonlinear Systems Identification via an Input-Output Model Based on a Feedforward Neural Network

ＮｏｎｌｉｎｅａｒＳｙｓｔｅｍｓＩｄｅｎｔｉｆｉｃａｔｉｏｎｖｉａａｎＩｎｐｕｔＯｕｔｐｕｔＭｏｄｅｌＢａｓｅｄｏｎａＦｅｄｆｏｒｗａｒｄＮｅｕｒａｌＮｅｔｗｏｒｋＯ．Ｌ．ＳｈｕａｉＳｏｕｔｈＣｈｉｎａＵｎｉｖｅｒｓｉｔｙｏｆＴｅｃｈｎｏｌｏｇｙ...