Instrumentation & Measurement Magazine 24-1 - 32

Metrological Aspects
In this section, the main aspects connected to measurement
systems, data processing and standards are discussed, and in
particular:
◗◗ Sensor synchronization in sensor network
◗◗ Calibration
◗◗ Traceability assurance
◗◗ Uncertainty assessment in big data processing
◗◗ Requirements of standard and of ISO EN 10012, in particular [12].

Sensor Synchronization in Sensor Networks
More and more networks of digital and low-cost sensors are
used, like digital MEMS sensors for measurement of different
quantities, such as temperature, vibration, force, etc.
Synchronous data are useful, and mostly in dynamic application, but phase of single measurement is difficult to
assess due to local analog to digital conversion and different
data rates. Synchronization should be also guaranteed in data
transmission by the protocol, and this aspect is not trivial [1].
As an example, the availability of networks of low-cost accelerometers makes possible modal analysis of large structures
for health monitoring, based on a large amount of measuring
points; the impact of time synchronization error should be corrected [17], [18], or, better, it should be prevented. Calibration
methods aiming at this purpose are now under study by National Metrological Institutes [19].

Calibration
International standards for calibration of mechanical and
thermal transducers in many cases consider calibration procedures based on reference instruments, whose output is an
analog signal [20], [21]; therefore, sometimes defining a " digital sensitivity " appears to be a quite new proposition, to be
harmonized in the standard itself [4].
New technologies propose sensors which are very low cost
and able to be embedded on the measuring point to locally
measure the quantity of interest, allowing a better monitoring
of the production process.
Mass production of low-cost sensors requires also that the
calibration cost is strongly reduced with respect to the procedure for traditional instrumentation.
Some proposal for mass " in line calibration " by the producers are suggested, but they are typically for static conditions,
and dynamic calibration seems to be difficult to realize [4]. The
issues that motivate the interest towards in-line calibration, to
be performed in dynamic conditions, have been extensively
investigated in [20], [21]. Authors in [19] confirm the need to
deepen these aspects, also by National Metrological Institutes.
Embedded sensors allow a closer interaction between control of the process and the process itself, and this situation is a
big opportunity. On the other hand, it is impractical or impossible to remove them from the measuring point for calibration
purposes [22].
Many proposals exist in literature to provide an on-line
calibration:
32	

◗◗ Transfer of the reference throughout the network: a reference signal could be transferred where the calibration
occurs (electrical signals) or the reference located near the
transducer to be calibrated could be activated at the due
time [13], [23];
◗◗ High performance devices are inserted into the network,
together with a large number of low-cost sensors, mainly
for calibration purposes; this configuration makes it
possible to remove the high-performance measurement
systems to guarantee traceability by means of laboratory
calibration [14];
◗◗ " Smart calibration " with careful monitoring of the
disturbing quantities by other sensors, improving precision and reducing bias effects; sometimes, this effect
could be improved by a suitable design of the measurement principle, reducing the possibility of bias [24];
◗◗ On site auto-calibration by a series of specific measurements, even though they are mainly valid for static
calibration.

Traceability
Traceability is a very important aspect, since it is universally
acknowledged that traceability of measurements is a mandatory requirement [4], [5], [11]. Most of the above described
techniques are helpful to realize the reproducibility of data,
therefore improving traceability of results.
To some authors, the possibility of carefully modelling and
monitoring the specific contribution of measurement uncertainty and errors will further improve traceability of results
[10], without limiting the acknowledgment of traceability only
to an unbroken chain of documented calibrations.

Data Processing
Big data processing and use of advanced algorithms like machine learning (ML) mainly impact the capability of modelling
uncertainty propagation throughout the whole process of data
analysis, including feature extraction, selection, ML application for classification and/or clustering [25]. Currently, the
main efforts are devoted to the identification of measurement
uncertainty models for sensors and sensor networks, including synchronization effects and digital data transmission.
This probabilistic information should be used to feed an uncertainty flow and propagation, which is now modelled by a
Monte Carlo simulation as a worthy starting attempt.
Nevertheless, a good idea is to try modeling some single
steps of the procedure, and reducing the uncertainty contributions as much as possible, in order to maintain the process
under control [1].

Updating Standards, Particularly ISO 10012
In the following section, general and specific aspects are summarized, with the aim of proposing an updated point of view
that will continue to guarantee the metrological requirements
of devices and measurement processes. At the same time, continuity should be assured with the basics of the requirements
for the measurement management system, considering also

IEEE Instrumentation & Measurement Magazine	

February 2021
Instrumentation & Measurement Magazine 24-1

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-1
