Productivity and Sustainability Management in the Responsive Factory

该项目的目标是利用工厂活动数据的组合来优化工业运营，并识别机会，提高效率、生产力和可持续性。数字传感技术的快速发展，使得在制造环境中实时记录活动既实用又实惠。然而，关于移动和活动的各种实时数据的可用性并不能自动帮助工程师管理现代工业操作中典型的复杂动态环境。要做到这一点，他们需要工具来支持他们以提高生产力和可持续性的方式解释不断变化的数据。换句话说，数字化制造带来的研究挑战不是数据的获取，而是缺乏计算方法来分析大量不同（即多模态）传感器数据，并识别允许工程师评估车间当前状态的模式，了解过去事件的影响，并预测事件对一系列措施的后果。在这种需求的推动下，以下提案详细介绍了一项工作计划，以调查是否可以将概率网络的形式扩展到处理工厂车间记录的多种形式的工业活动数据，这些网络形式已被用于从其他环境中的位置跟踪数据（例如交通模型中的车辆运动和家庭环境中的个人日常活动）中生成计算模型。这样一个模型将允许使用制造活动的各种信号（例如，材料运输、工作人员移动、振动、电流和空气质量等）来推断工业工作场所的行为，并产生量化措施，以支持影响场地生产和可持续性绩效的决策。

The ambition of this project is to use a mix of factory activity data to optimise industrial operations, and to identify opportunities and deliver improvements in efficiency, productivity and sustainability. The rapid advance of digital sensing technologies, is making the real time recording of activities in a manufacturing environment both practical and affordable. However, the availability of diverse, real time data about movement and activity does not automatically help engineers manage the complex, dynamic environments typical of modern industrial operations. To do this they need tools that support their interpretation of constantly changing data in ways that enhance productivity and sustainability. In other words, the research challenge posed by digital manufacturing is not the capture of data, but rather the lack of computational methods to analyse large flows of diverse (i.e. multimodal) sensor data and recognise the patterns that allow engineers to assess the current state of the shop floor, understand the impact of past events and predict the consequences of incidents on a range measures. Motivated by this need, the following proposal details a program of work to investigate if the forms of probabilistic networks that have been employed to generate computational models from location tracking data in other contexts (e.g. vehicles movements in traffic models and the daily routines of individuals in domestic environments) can be extended to work with multiple forms of industrial activity data recorded on a factory floor. Such a model would allow diverse signals of manufacturing activity (e.g. material transport, staff movement, vibration, electrical current and air quality etc.) to be used to infer the behaviour of an industrial workplace and generate quantitative measures that support decisions which impact on a sites' production and sustainability performance.