Mixture model based batch process monitoring and fault diagnosis for sustainable manufacturing

基于混合模型的可持续制造批量过程监控和故障诊断

• 批准号: 445520-2012
• 负责人: Yu, Jie
• 金额: $ 1.3万
• 依托单位: McMaster University
• 依托单位国家: 加拿大
• 项目类别: Collaborative Research and Development Grants
• 财政年份: 2012
• 资助国家: 加拿大
• 起止时间: 2012-01-01 至 2013-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=498295
• 关键词: Mixture; model; based; batch; process

Industrial manufacturing is facing significant challenges on sustainable resources, energy, environment and profits. Batch or semi-batch processes have been widely applied to different industries including energy, chemical, materials, semiconductor, pharmaceutical, biotechnology and food industries for producing high-value-added products. Therefore, it is critically important to develop reliable monitoring systems of batch processes with strong fault detection and diagnosis capabilities in order to ensure safe, profitable and environmentally sustainable operation in complex industrial manufacturing. Typically batch processes have inherent features including multiplicity of operating phases, phase based shifting dynamics, between-phase transient behaviors, process nonlinearity and non-Gaussianity, batch-to-batch variations, and uncertainty of phase divisions. All these characteristics constitute significant challenges for effective batch process monitoring and diagnosis in industrial practice. This research project is aimed to develop the novel kernel mixture model and Bayesian statistics based smart monitoring techniques for abnormal event detection, fault propagation identification and root cause diagnosis of multiphase batch processes. With the adaptive monitoring and diagnosis approaches, multiphase batch processes can be automatically maintained and optimized with the best safety, productivity, quality and energy efficiency as well as the lowest environmental emissions and carbon footprints. The developed technology will be demonstrated in manufacturing facilities at Dow Chemical and further be extended to different industries as generic technical solutions. It is anticipated that the research findings and inventions of this project will benefit a wide range of industries in Canada with significant economic, environmental and social values such as improved productivity, increased energy efficiency, reduced carbon emissions and mitigated safety incidents. Moreover, this research will make substantial knowledge contributions to chemical engineering and particularly process systems engineering fields with new methodologies on process monitoring, fault diagnosis and process sustainability.

工业制造业正面临着可持续资源、能源、环境和利润的重大挑战。间歇或半间歇工艺已广泛应用于不同行业，包括能源、化工、材料、半导体、制药、生物技术和食品工业，以生产高附加值产品。因此，开发具有强大故障检测和诊断能力的可靠的间歇过程监控系统，以确保复杂工业制造中安全，有利可图和环境可持续的运行至关重要。典型地，间歇过程具有固有特征，包括操作阶段的多样性、基于相的移动动态、相间瞬态行为、过程非线性和非高斯性、批次间变化以及相划分的不确定性。所有这些特点构成了在工业实践中有效的间歇过程监测和诊断的重大挑战。本研究旨在发展一种新的基于核混合模型和贝叶斯统计的智能监控技术，用于多相间歇过程的异常事件检测、故障传播识别和根本原因诊断。通过自适应监测和诊断方法，可以自动维护和优化多相批处理过程，实现最佳的安全性、生产率、质量和能源效率，以及最低的环境排放和碳足迹。开发的技术将在陶氏化学的生产设施中进行演示，并作为通用技术解决方案进一步扩展到不同行业。预计该项目的研究成果和发明将惠及加拿大的多个行业，具有显著的经济、环境和社会价值，如提高生产力、提高能源效率、减少碳排放和减少安全事故。此外，这项研究将通过过程监测、故障诊断和过程可持续性的新方法，为化学工程，特别是过程系统工程领域做出实质性的知识贡献。