Bayesian Network Based Approach for Diagnosis of Modified Sequencing Batch Reactor

Wastewater treatment is a complicated dynamic process affected by microbial, chemical and physical factors. Faults are inevitable during the operation of modified sequencing batch reactors(MSBRs) because of the uncertainty of various factors. Abnormal MSBR results require fault diagnosis to determine the cause of failure and implement appropriate measures to adjust system operations. Bayesian network(BN) is a powerful knowledge representation tool that deals explicitly with uncertainty. A BN-based approach to diagnosing wastewater treatment systems based on MSBR is developed in this study. The network is constructed using the knowledge derived from literature and elicited from experts, and it is parametrized using independent data from a pilot test.A one-year pilot study is conducted to verify the diagnostic analysis. The proposed model is reasonable, and the diagnosis results are accurate. This approach can be applied with minimal modifications to other types of wastewater treatment plants.     

基于油液监测的船舶柴油机故障预测与健康管理技术研究

以连续监测收集的油液光谱数据为依据,提出了基于统计过程控制（SPC）的船舶柴油机故障预测方法,将主要磨损元素的质量浓度梯度作为质量特性指标,运用SPC技术对各质量特性进行了统计分析.结果表明,此船舶柴油机可能存在异常因素而导致磨损状态处在统计控制之外,需要查找故障的根源;并基于上述获得的油液监测数据,利用投影寻踪（PP）方法对此船舶柴油机整机性能退化程度进行了综合评定,制定了相应的评判标准.     

沥青混合料拌和过程实时监控

为控制道路施工过程中沥青混合料的拌和质量与拌和状态, 提出一种以非介入方式利用模板匹配识别算法实时提取骨料、粉料、沥青质量数据、拌和时间及温度等沥青混合料主成分数据信息的方法, 根据识别到的沥青混合料数据信息建立了数据采集与传输的时序逻辑关系; 在WEB监控中心下可视化显示了沥青混合料配合比误差、级配误差、拌和时间和温度等关键信息, 并利用这些多模态信息融合策略评价了沥青混合料的拌和质量; 根据施工过程中沥青混合料类型的先验知识分析了混合料数据的动态变化, 在无人工干预的情况下自动识别了实时生产的沥青混合料类型; 建立了骨料数据的模型分布, 并结合拌和时间判断拌和设备的运行和筛分状态; 存储实时接收到的数据, 实现了沥青混合料历史数据跨时间查询和成本评判。研究结果表明: 利用模板匹配识别算法采集沥青混合料字符数据时间为4.9 ms, 识别准确率达100%, 满足了施工中沥青混合料拌和数据采集时间间隔小于0.02 s的要求, 实现了施工过程中沥青混合料数据的连续检测、自动识别、实时跟踪和可视化监控; 当沥青混合料质量不合格或拌和设备出现故障时可实时预警, 为综合评价沥青混合料拌和过程与实时掌控沥青混合料拌和质量提供了依据。      

On-line Batch Process Monitoring and Diagnosing Based on Fisher Discriminant Analysis

Introduction Nowadays, it has been paying more attentionto the batch or semi-batch process which producedthe high added value and high quality products inchemical industry, such as polymer, pharmaceuti-cal and semiconductor industry.The aim of the batch process monitoring is tokeep the product quality uniform and consistentunder batch-to-batch process and meet specifica-tion limits. The direct approach is to monitor cru-cial product quality variables in batch process.However, quality variables…     

Memory based scheme to monitor non-random small shift patterns in manufacturing process

Consistent high-quality and defect-free production is the demand of the day. The product recall not only increases engineering and manufacturing cost but also affects the quality and the reliability of the product in the eye of users. The monitoring and improvement of a manufacturing process are the strength of statistical process control. In this article we propose a process monitoring memory-based scheme for continuous data under the assumption of normality to detect small non-random shift patterns in any manufacturing or service process.The control limits for the proposed scheme are constructed. The in-control and out-of-control average run length(AVL) expressions have been derived for the performance evaluation of the proposed scheme. Robustness to non-normality has been tested after simulation study of the run length distribution of the proposed scheme, and the comparisons with Shewhart and exponentially weighted moving average(EWMA) schemes are presented for various gamma and t-distributions. The proposed scheme is effective and attractive as it has one design parameter which differentiates it from the traditional schemes. Finally, some suggestions and recommendations are made for the future work.     

Modeling of additive manufacturing process relevant feature in layer based manufacturing process planning

Compared with general machining processes, additive manufacturing (AM) process has stabler planning route and limited process variables and this makes it to be more easily designed and planned with knowledge based systems and computer aided techniques. Case based reasoning approach is applied to the process planning of additive manufacturing in this paper. The concept of “AM process relevant design features” is proposed after the analysis of the characteristics of AM processes. The concept is used as the basis of the knowledge representation, and AM relevant feature graph is used as the case representation schema. The case retrieval method is discussed based on this graph. The case representation of a machine arm is given to illustrate the brief process of the proposed approach.     

Bayesian reliability assessment and degradation modeling with calibrations and random failure threshold

A degradation model with a random failure threshold is presented for the assessment of reliability by the Bayesian approach. This model is different from others in that the degradation process is proceeding under pre-specified periodical calibrations. And here a random threshold distribution instead of a constant threshold which is difficult to determine in practice is used. The system reliability is defined as the probability that the degradation signals do not exceed the random threshold. Based on the posterior distribution estimates of degradation performance, two models for Bayesian reliability assessments are presented in terms of the degradation performance and the distribution of random failure threshold. The methods proposed in this paper are very useful and practical for multi-stage system with uncertain failure threshold. This study perfects the degradation modeling approaches and plays an important role in the remaining useful life estimation and maintenance decision making.     

传感器性能退化与结构损伤响应异常区分研究

针对现行结构损伤检测方法仅考虑了环境和荷载对损伤的影响,未考虑传感器因时间和环境变化导致的性能退化从而严重影响损伤诊断率的问题,考虑到传感器性能退化的局部性及结构损伤的全局性,利用最小均方差估计与广义极似然比（GLRT）理论检测结构响应异常,结合统计质量控制图的控制指标超出界限的波动样式判断异常来源,并通过简支梁数值算例验证了本文提出的方法.研究结果表明:利用广义极似然比可以有效检测响应异常,利用控制图可以辨别响应异常来源;传感器性能退化导致控制指标超出界限的波动为非平稳波动,而损伤导致控制指标在一定范围内的波动为平稳波动.      

基于贝叶斯网络的CTCS-3级列控车载系统韧性

为弥补现有指标的不足，引入韧性作为非常态事件下CTCS-3级（China train control system-3）列控车载子系统运行稳定性的测度指标. 提出了车载子系统韧性量化评估方法，构建了基于贝叶斯网络（Bayesian network, BN）的韧性评估模型，并定义了5种基于韧性的部件重要度指标；进一步利用贝叶斯网络双向推理功能，计算了车载子系统在不同扰动情景下的韧性及部件重要度指标. 研究结果表明:韧性可全面描述车载子系统抵御扰动和从扰动中恢复的能力，非常态事件扰动下，韧性与可用性指标存在明显差异；不同扰动情景下系统韧性明显不同，扰动发生时，车载子系统面临磁暴影响时的韧性为0.8017，而遭遇雷电时的韧性为0.8819，面临冰雪扰动时的韧性为0.9880；部件重要度存在情景依赖，同一部件在不同扰动情景下重要度排序可能不同，且可能随时间动态变化.      

Residual Chart with Hidden Markov Model to Monitoring the Auto-Correlated Processes

Autocorrelations exist in real production extensively, and special statistical tools are needed for process monitoring. Residual charts based on autoregressive integrated moving average (ARIMA) models are typically used. However, ARIMA models need a quite amount of experience, which sometimes causes inconveniences in the implementation. With a good performance under less experience or even none, hidden Markov models (HMMs) were proposed. Since ARIMA models have many different performances in positive and negative autocorrelations, it is interesting and essential to study how HMMs affect the performances of residual charts in opposite autocorrelations, which has not been studied yet. Therefore, we extend HMMs to negatively auto-correlated observations. The cross-validation method is used to select the relatively optimal state number. The experiment results show that HMMs are more stable than Auto-Regressive of order one (AR(1) models) in both cases of positive and negative autocorrelations. For detecting abnormalities, the performance of HMMs approach is much better than AR(1) models under positive autocorrelations while under negative autocorrelations both methods have similar performances.     

高速公路尾随相撞事件的贝叶斯网络致因分析模型

应用上海市高速公路1104条事件数据,基于专家知识和数据融合方法建立贝叶斯网络结构;利用服从Dirichlet分布的贝叶斯方法进行参数学习;运用团树传播算法进行推理分析。研究了上海市高速公路尾随相撞事件类型与不同道路环境条件之间的关系。在验证贝叶斯网络模型的有效性后,系统分析事件致因,并提出改进措施。发现重大尾随相撞事件易发生在大中型车与小型车之间;夜间易发生大中型货车的重大尾随相撞事件,尤其是凌晨0时至6时;路表潮湿状态下的非普通路段上易发生大中型客车的重大尾随相撞事件。结果表明贝叶斯网络建模能够更好的反映事件致因因素的多维性及关联性,是交通事件致因分析的有效方法。     

一种基于激光与视频信息时空数据融合的行人检测方法

针对城市交通行人安全问题,本文提出了一种基于激光与视频数据融合的行人检测方法.通过激光与视频数据空间和时间上的融合,将激光数据映射到图像坐标;在激光聚类过程中,采用K-means 聚类算法对激光云点进行聚类分析,然后运用行人宽度模型提取候选行人区域;在基于图像的行人检测过程中,选取头肩、躯干以及腿部人体特征部位,采用Haar-like 特征集和Boosting 算法进行训练,得到部位检测器;最后,基于贝叶斯决策的组合策略对候选行人区域进行有效判定.实验结果表明,本文所述算法有较好的检测精度和实时性能.     

基于贝叶斯网络组合模型的公交驻站时间预测

公交驻站时间是公交行程时间的主要组成部分,其预测精度直接影响智能公交系统中公交信息发布的准确性.为了提高公交驻站时间的预测精度,提出一种基于贝叶斯网络的组合预测模型,它由反向传播神经网络和径向基函数神经网络模型组成.首先利用两种神经网络模型预测公交驻站时间;然后利用改进后的等宽数据离散方法,将两种神经网络的预测结果和观测的驻站时间数据离散后用于贝叶斯网络学习;最后通过贝叶斯网络推理得到驻站时间组合预测结果.实例分析表明,贝叶斯网络组合模型驻站时间预测结果的误差指标均优于单一模型,证明其可有效提高单一模型的预测精度.     

民航运输航空器着陆阶段偏出跑道事件分析模型

为了克服故障树分析方法只能描述正常和故障两种状态的缺陷,利用故障树方法的逻辑分析优势,以及贝叶斯网络描述多态事件和计算概率的功能,建立了基于故障树和贝叶斯网络相融合的航空器着陆偏出跑道事件分析模型,提出了故障树与贝叶斯网络之间的转换算法求解该模型.根据1996—2010年中国民航着陆偏出跑道事件的数据,确定了着陆航空器偏出跑道的主要原因,并按重要程度进行了排序.研究结果表明:积水、反喷或减速板故障、复杂气象条件、驾驶术欠缺、前轮转弯卡阻、夜航或受灯光不利影响、对机组资源管理失效、积冰积雪是导致着陆偏出跑道的主要风险因素,其后验概率均大于0.3;应针对主要风险因素,制定针对性预防措施.      

基于改进SBELM的耦合故障诊断方法

为了对主减速器的耦合故障进行识别,通过对振动信号经过集成经验模态分解(ensemble empirical mode decomposition, EEMD)所获得的高频分量采用自适应阈值降噪和对低频分量采用区间阈值降噪,有效去除了信号噪声,创建了配对多标签分类策略(paired multi-label classification,PMLC).基于PMLC和稀疏贝叶斯极限学习机(sparse Bayesian extreme learning machine, SBELM)用单故障样本构造概率分类器集,再采用网格搜索方法生成最优决策阈值,将分类器集的概率输出转换为耦合故障模式,提出了基于自适应区间阈值降噪和SBELM的耦合故障诊断方法,并用主减速器的实际样本集验证了该方法的性能.研究结果表明:该方法的诊断精确度达到96.1%,比基于PNN(probability neural networks)和SVM(support vector machine)的诊断方法提高了5%;该方法的训练时间和执行时间为131.4和61.3 ms,比基于SVM的诊断方法减少了70%.      

国内外实时交通流数据质量控制比较与分析

交通流数据质量控制可以保证从数据源所获得数据的正确性和完整性，并为数据的管理和应用提供可靠的数据基础，虽然历史数据修正方法和交通流理论修正方法能够用于质量控制，但难以实现推广应用。根本原因在于前者需要大量历史数据，引发了存储上的困难；后者需要针对不同路段进行重复建模，模型通用性差，且准确性难以评价，鉴于此，本文首先研究并设计了基于线性插值法的道路交通流数据质量控制方法，该方法主要以时间点邻近数据为参考，对交通流参数和时间点分别进行判别修正，其优点在于不需要大量的历史数据，且适用性和可操作性强，其次，本文建立了实时交通流数据质量控制平台，针对北京市和美国圣安东尼奥市的数据进行质量控制处理，归纳总结出国内外实时交通流数据在质量控制前后的特征及差异，证明该方法能够有效的解决数据质量问题，最后，本文针对实时交通流数据质量控制方法的选取、交通探测器选择和配置等方面提出了建议。     

Research on online texture measurements in metal rolling field

In rolled strip material, the orientation of the crystallites, known as texture, is influenced by various kinds of thermo-mechanical processes, such as casting, plastic deformation, annealing and phase transformation. The modern industry production requires stable product performance, real-time monitoring and full controlling of the quality. The online texture measurement in metal rolling can be used to real-time monitor the whole process, and then feedback control to the production process can be implied to adjust the process parameters to ensure the stability of the products. The principles, advantages and disadvantages of related detection methods (2D X-ray diffraction, neutron diffraction, laser-ultrasonics and electromagnetic acoustic transducers (EMAT)) and the possibility of online measurement are discussed. Finally, 2D X-ray diffraction and laser-ultrasonics are employed on online texture measurement, and the schemes of online texture measurement are proposed.     

Weld Geometry Monitoring for Metal Inert Gas Welding Process with Galvanized Steel Plates Using Bayesian Network

We present a novel method to monitor the weld geometry for metal inert gas(MIG)welding process with galvanized steel plates using Bayesian network(BN),and propo...     

基于集成神经网络的刀具磨损量监测

提出了一种基于集成神经网络识别铣刀磨损量的监测方法．利用小波包变换将切削力和振动信号分解为不同频带的时间序列，从每个信号中选择与刀具磨损状态最相关的3组频段的均方根作为监测特征；通过信号的组合和不同子网络输出决策间的融合，集成神经网络输出刀具磨损的识别结果．试验和仿真分析表明，此方法能够满足刀具磨损量实时监测的要求．     

基于加权小波子带图像的贝叶斯人脸识别算法

对贝叶斯人脸识别公式进行了简化,在此基础上设计了基于加权小波子带图像的贝叶斯人脸识别算法.首先对人脸图像进行小波分解,把分解得到的低频子图与类内均值做差作为类内差异图像进行贝叶斯测试,选择相似度最高的N幅图像作为候选图像,然后对候选图像再次利用高频子图与对应频段的类内均值做差作为模式矢量并行进行贝叶斯测试,通过加权排序得到最后结果.实验结果验证了该方法的有效性.