Objective 1: To develop calibration methods for industrial sensors of dynamic measurements such as acceleration, force and pressure with digital data output (data streams) and internal digital pre-processing, including the extrapolation of the measurement uncertainty from individually calibrated sensors to other individuals of the same type by means of co-calibration and statistical modelling.
The project has developed the required basic concepts, terminology and specifications for calibration methods for industrial sensors of dynamic measurements. As a first step, the project developed a micro-controller (µC) board that can hold digital sensors and provide time-stamping traceable to the SI. A communication interface based on Protobuf-Messages connects the board to upstream systems. These are currently running the agentbased framework (ABF) developed within the project to enable easy integration with data analysis method developments. The ABF in turn, is capable of providing the measurement information in arbitrary protocols. For example, the OPC-Unified Architecture was implemented, which provides data streams from the board to connected PCs and will later be used for the implementation in the project’s testbeds. For the communication of the data from the µC board, a collaboration with the EMPIR project 17IND02 SmartCom has been initiated to develop a joint demonstrator for metrology in the FoF. The software corresponding to the µC board and the adaption to the agent-based framework has been published on the project’s open source repository on GitHub (https://github.com/Met4FoF).
A first application of the new hardware was as a dynamic calibration concept as an extension to the existing acceleration calibration facilities at PTB. As part of this, the µC board provides an additional ADC channel which samples (and timestamps) a dedicated synchronisation signal in parallel to the DUT’s digital output. First prototypical phase response measurements were successfully performed with this extension.
Objective 2: To develop and demonstrate methods enabling digital sensors to provide uncertainty and/or data quality information together with the measurement data.
The project started development of a proof-of-concept “Smart Traceability” sensor by extending a conventional sensor with a “Smart-up Unit”, such that it provides measured values together with their associated uncertainty and other relevant data quality information. Furthermore, the project further developed the software from EMPIR project 14SIP08 by extending the corresponding software library PyDynamic. The existing PyDynamic software library was extended to a continuous integration (CI) workflow for automated software quality assurance. The central software repository on GitHub connects PyDynamic and the other software developments from this project to an implementation of the ABF developed in objective 1. In addition, the development of modules for reading and writing digital calibration certificates has been started in collaboration with 17IND02 SmartCom and the joint Stakeholder Advisory Board (SAB) for the projects.
For the extension of the ZEMA and STRATH testbed, the prototype “Smart-up Unit” is being prepared to integrate MEMS sensors measuring temperature and acceleration.
Objective 3: To develop a cost-efficient in-situ calibration framework for MEMS sensors measuring ambient temperature for their integration into an industrial sensor network with metrological quality infrastructure.
The project has begun the development of an automated testbed equipment (ATE) at the SPEA testbed for traceable in-situ calibration of MEMS temperature sensors using an optimised network of reference sensors in an automated test environment. The initial design of the ATE setup has been drafted and development of the corresponding parts has started. An initial version of the laboratory calibration facility for the traceable calibration of the on-board temperature sensors of the reference fixture has been finished. First data analysis and FEM simulations have been carried out and will be used to further optimise this novel calibration setup; that uses on-board temperature sensors to generate a temperature mapping for the calibration of the MEMS sensors. In addition, together with 17IND02 SmartCom the calibration facility is developed into a demonstrator for the automated creation of digital calibration certificates.
Objective 4: To develop and assess data aggregation methods for industrial sensor networks based on machine learning and efficient software architectures, addressing synchronisation of measurements, making use of redundancies of measurements, taking into account uncertainty from calibration and network communication issues, including strategies for balancing cost versus uncertainty and explore methods to identify the measurement coverage and accuracy required for process output targets.
Generic mathematical models have been derived and potential uncertainty evaluation approaches have been identified by the project. In particular, typical methods for feature extraction methods as the first step in ML have been extended with uncertainty propagation. Several approaches for the assessment and exploitation of redundancy in sensor networks have been identified and applied to testbed data sets. Based on models of the data collected by digital sensors (subject to noise and jitter effects), methods have been developed to account for timing and synchronisation issues. With the implementation of these methods in an ABF, the flexible use of a variety of sensor networks can be achieved. The basic outline of the ABF has been created and its implementation started. The plan is to initially use the ABF to simulate a heterogenous sensor network.
Measurement analysis for sensor networks in the FoF is typically based on ML methods for decision-making. In order to bring metrology into this area, the uncertainty associated with the raw data streams has to be taken into account by the ML methods applied. As a first step, the methods already in use for the STRATH and ZEMA testbed data are being investigated regarding their extension to take into account measurement uncertainties. In addition, probabilistic ML methods are being evaluated with the data sets so far provided by the project’s testbeds. In order to increase early up-take of these developments, initial data sets have been made publicly available on the Zenodo repository under a dedicated Met4FoF community page (https://zenodo.org/communities/met4fof/. In addition, the development of web-based tutorial for using the data sets has been started. A first version of a tutorial webinar on the ABF system has also been recorded and published. Web-based tutorials on the available data sets have been integrated with the project’s GitHub repository and three data sets have been published with open access: STRATH data set, ZeMA data set of one cylinder, ZeMA data set with three cylinders. The data sets will be updated and combined with further web-based tutorials on their application with Python-based ML taking into account measurement uncertainties.