Sensors Attributes. The statistics available to the optimizer depend on how the sensors are embedded in the in- dustrial process. For example, in many WSN architectures multiple sensors are placed in close proximity to each other measuring similar attributes of the industrial process. Alternatively, there may be only one sensor reading a specific process. A general distinction can be made between data acquired from a single sensor as opposed to multiple sensors. In either case statistics may be requested on either a single or multiple attributes. Attributes may include: measured sensor values, location (xy coordinates), timestamps, etc. So in a generic setting, the optimizer may be able to specify attributes for:
1. Single Sensor
a. Single Attribute
b. Multiple Attribute
2. Multiple Sensors
a. Single Attribute
b. Multiple Attribute H2020-SPIRE-2014 DISIRE The purpose of this document was to review our findings from the industrial process discovery stage of our WP and describe the prototype. We also see this document as an opportunity for us to introduce our WP to our other DISIRE partners and we fully encourage our partners’ further enquiries and cooperation. We started off by giving some examples of the typical uses of sensors in the industry which included control and monitoring functionality as well as for safety purposes. Based on what we have learned from our DISIRE partners, we defined a basic typology for industrial sensory output from a statistical analysis perspective: big data vs. small data, and real-time vs. batch. With these preliminary distinctions in place, we then went on to demonstrate the different parts of our WP implemented through a prototype workflow model. The first part of the prototype deals with the technical aspects of collecting and organizing the outputs from the sensors em- bedded in the industrial process. This includes reading data files, cleaning and QA-ing the data for erroneous readings, and preparing the data to be stored in the sensor databases for further manipulation in the statistical engine. The second part deals with the statistical engine module which is tasked for preparing the data for the optimal control module. This includes performing various tasks on the data such as basic statistics, clustering and segmentation, aggregation, modeling and forecasting, outlier analysis and dimensionality reduction, and more. As was stressed in the document, throughout our WP we will explore which technologies work best in performing these tasks ...
