Clustering and dimensionality reduction based techniques for low latency big data feature extraction applications

用于低延迟大数据特征提取应用的基于聚类和降维的技术

• 批准号: 522286-2017
• 负责人: AlAnbagi, Irfan
• 金额: $ 1.82万
• 依托单位: University of Regina
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=640689
• 关键词: Clustering; dimensionality; reduction; based; techniques

The Mercedes-Benz Fuel Cell Division (MBFC) in Burnaby, Canada develops and runs the manufacturingprocesses required for the assembly of Fuel Cell Stack prototypes. MBFC uses the Manufacturing ExecutionSystem (MES) to collect and analyze data from the manufacturing lines to the database system. The size of thecollected data is very high and MBFC is not able to detect certain fuel cell defects in a timely manner.This Engage grant project aims to develop and evaluate a mechanism that reduces the time taken to search theMercedes-Benz Fuel Cell (MBFC's) big data and detect a failure in the manufacturing process in a real-timefashion. This will be done by critically studying and understanding MBFC's database schema and analysing agraph version of this database. A dimensionality reduction and feature extraction mechanism will be developedto search a specific cluster of the MBFC big data set and identify defective fuel cells using the Gas DiffusionLayer (GDL) bleed through method. Finally, a mechanism will be proposed to detect defective fuel cells in atotal data dump scenario.The proposed research is innovative because it will devise a novel real-time big data feature extractiontechnique that combines a low latency and high efficiency dimensionality reduction technique with a clusteringgraph-based database mechanism. Potential pitfalls and risks will be identified, mitigated and managed throughenhanced research methodologies, using diverse dimensionality reduction models and providing alternativeapproaches to detect features in big data.MBFC will be involved in the project by providing technical support, consultation and advice on issuesrelated to their database schema, Manufacturing Execution Systems (MES) interface and their GDL bleedthrough testing processes. This research will enable MBFC to establish a presence in fuel cell manufacturingand help establish Canada as a global leader in innovative green technologies.

位于加拿大本拿比的梅赛德斯-奔驰燃料电池部（MBFC）开发并运行燃料电池堆原型组装所需的制造流程。MBFC使用制造执行系统（MES）来收集和分析从生产线到数据库系统的数据。收集到的数据量非常大，MBFC无法及时发现某些燃料电池缺陷。该Engage资助项目旨在开发和评估一种机制，以减少搜索梅赛德斯-奔驰燃料电池（MBFC）大数据所需的时间，并实时检测制造过程中的故障。这将通过批判性地研究和理解MBFC的数据库模式和分析该数据库的agreeable版本来完成。将开发一种降维和特征提取机制，以搜索MBFC大数据集的特定聚类，并使用气体扩散层（GDL）渗透方法识别有缺陷的燃料电池。最后，本文提出了一种在数据转储场景下检测缺陷燃料电池的机制，该机制将低延迟、高效率的降维技术与基于聚类图的数据库机制相结合，提出了一种新的实时大数据特征提取技术。通过增强研究方法，使用不同的降维模型并提供检测大数据特征的替代方法，将识别、缓解和管理潜在的陷阱和风险。MBFC将参与该项目，就与其数据库模式、制造执行系统（MES）接口和GDL渗透测试流程相关的问题提供技术支持、咨询和建议。这项研究将使MBFC能够在燃料电池制造业中占有一席之地，并帮助加拿大成为创新绿色技术的全球领导者。