Transformation of H&S Real-time Sensors Data into Information and Knowledge: Experiences, Future Needs, and Applied Processes

Posted on by Emanuele Cauda, PhD; Eelco Kuijpers, PhD; and Jean-Philippe Gorce, PhD

 

Data is the oil of the 21st century [1] and the key component of the fourth industrial revolution. Data will affect every aspect of life, including the workplace. In the workplace, sensors used for production, optimization, logistic, quality control, and health and safety are the among the largest contributors of data. Like oil, data is a raw material, and it needs to be harvested, transformed and processed to deliver higher value and its true impact.

The NIOSH Center for Direct Reading and Sensor Technologies is engaged in activities related to the data generated by health and safety sensors and the critical step of transforming this data into applicable information and new knowledge. The proposed “right sensors used right” approach is part of these activities. Time-resolved sensors are sensors that generate signals and data in time. The interest in these technologies to be used in the workplace for health and safety is increasing exponentially thanks to the miniaturization of the components, the potential of the sensors, and the capability of transferring, storing, and processing data streams at high speed. These time-resolved monitors do and will generate a substantial and continuous flow of data. The success of these technologies lies in the ability to transform the data into actionable information that can reduce exposures and ensure safe working conditions (i.e. through early warning systems or more data driven intervention strategies). The scientific community is called upon to lead this transformation through the development of applied approaches and methodologies to facilitate it.

The NIOSH Center is honored to collaborate with the Netherlands Organisation for Applied Scientific Research (TNO) and the Health and Safety Executive (HSE) in Great Britain. NIOSH, TNO and HSE have a strategic collaboration which includes current research on the application of sensors for enhancing occupational health and safety. Examples of products from this partnership include:

  • the development of guidelines for the calibration and validation of particulate matter sensors [2] in laboratory and field settings and
  • the deployment of PM sensors in occupational settings aimed at developing methods for the collection and interpretation of sensor data for occupational exposures.

 

Additionally, a working group of the three organizations recently conducted an internal workshop titled: “Transformation of H&S real-time sensors data into information and knowledge: experiences, future needs, and applied processes”. During the workshop, the group discussed four different aspects of the transformation process (needs, models, holistic, and users) and the most important considerations are reported in this blog.

Needs

JP Gorce (HSE) and Maaike Le Feber (TNO) led the group in a discussion of the need for contextual data and data infrastructure as a first step in the transformation process. Location information can be used to complement primary sensor data, (i.e. bring context to exposure data), or alternatively be used on their own as primary data, (e.g. as a proxy for risk of personal exposure to hazards present in the workplace). Jobs, tasks, production activity, and operational data are other examples of contextual data to consider. Obtaining this data requires systematically thinking through the process of exposure generation all the way to exposure uptake. The identification of the overall objective is critical since the collection of each contextual data entails investment of resources and time. Finally, a more systematic approach for the collection of (contextual) data can increase the acceptability from the workers. The group discussed the importance of an online data infrastructure system (e.g. like EXCITE). An ideal system should adopt systematic approaches for the transformation but at the same time it should be modular and adaptable to new needs and experiences. An ideal system should accept sensors data and contextual non-sensor data simultaneously. The system should facilitate the transformation of data into different levels of information – both intermediate and final – based on the users’ need.

Models

Peter Wessels (TNO) and Abbas Virji (NIOSH) transitioned the workshop towards the discussion on the advanced mathematical and statistical models needed for transforming the data into information. They presented a case of dynamic noise mapping and the application of classification, unsupervised clustering, and then prediction. They discussed the need for models to accept challenges and constant tuning and challenged the concept of “real-time” arguing that for some applications, a 5–10-minute window average can be sufficient. The group agreed that “time-resolved” data is a more appropriate terminology for sensors’ measuring or averaging time interval to avoid the misconception of “real-time”. The benefit of using Bayesian spline models to time-resolved data that accounts for certain data characteristics not simultaneously addressed in standard statistical packages was discussed. There is a need to invest time and resources in developing, training, and tuning the models. The adoption of this and other statistical models to sensors data provides users with summary statistics including their probability but will require the user to think in terms of probability of a certain event to occur (for example overexposure).

Holistic

Emma Bushell and Miles Burger (HSE), Imelda Wong (NIOSH), and Henk Goede (TNO) discussed the consideration of data from sensors and the associated information as a component of occupational health and safety frameworks – frameworks that must consider the workers themselves as a central piece. Any entity interested in adopting sensors and generating data about workers with sensors must invest time and resources in educating workers via surveys, establishing ambassadors, and accepting the presence of boundaries. Fatigue detection technologies were discussed including the role of user acceptance/implementation in addition to and performance. Further information about choosing and implementing fatigue detection devices has been addressed in prior NIOSH Science Blogs (Choosing the “Right” Fatigue Monitoring and Detection Technology and The Who, What, How and When of Implementing Fatigue Monitoring and Detection Technologies). The decision should include an analysis of the benefits and limitations of the data and associated information. Hybrid and combined solutions from sensors and non-sensors data will be an option in the future with the side risk of conflicting information. The benefit of using data and information from sensors to foster positive behavioral change in workers was also discussed. If the behavior is considered a hazardous condition, a revised NIOSH hierarchy of control can be adopted, and sensors can be helpful tools.

Users

Kevin Dunn (NIOSH) and Paul Smith (HSE) presented the perspective of two experienced users on the adoption of sensors for occupational health and safety and the transformation to information and knowledge. The benefit of using time-resolved sensors to address questions that otherwise do not have answers for nanoparticles exposure were explained and the importance of detailed observation and notes of the process on which to base professional judgement in the transformation of data to information and knowledge were stressed. It was proposed that sensors data be used in two distinct phases. First, sensors data can be used for an in-depth evaluation for understanding a process and environment and controlling emissions and exposures. Second, the sensors can be used to monitor the process for unexpected emissions. The idea of direct parameters vs. indirect parameters was also discussed. Direct parameters can be related to compliance status within a certain regulation while indirect parameters can be used as leading indicators or process metrics. In situations where compliance does not exist, the overall community needs to work outside the idea of compliance with data from sensors.

Final Considerations

NIOSH, TNO, and HSE wrapped up the workshop with the commitment to continue working and interacting in this field. The three organizations understand that there are several aspects that need to be considered for this transformation including diversified data and their management, the mathematical and statistical models applied, the engagement of workers, and the need to create frameworks different from the strictly regulatory ones. The organizations welcome the feedback and comments of stakeholders on these important aspects.

 

Emanuele Cauda, PhD, is a Research Engineer in the NIOSH Pittsburgh Mining Research Division and co-Director for the NIOSH Center for Direct Reading and Sensor Technologies.

Eelco Kuijpers, PhD, is a scientist innovator at TNO, the Netherlands, involved in the application of sensors at the workplace and co-leading the VOHA proposition.

Jean-Philippe Gorce, PhD, is a scientist at the HSE Science and Research Centre based in Buxton, Great Britain, leading projects on the use of wireless sensors and personal positioning technologies applied to workplace health and safety.

 

References

  1.  Arthur C [2013]. Tech giants may be huge, but nothing matches big data. The Guardian.
  2. Ruiter S, Kuijpers E, Saunders J, Snawder J, Warren N, Gorce JP, Blom M, Krone T, Bard D, Pronk A, Cauda E [2020]. Exploring evaluation variables for low-cost particulate matter monitors to assess occupational exposure. International Journal of Environmental Research and Public Health 17(22):1-18 DOI: 10.3390/ijerph17228602.
Posted on by Emanuele Cauda, PhD; Eelco Kuijpers, PhD; and Jean-Philippe Gorce, PhD

Post a Comment

Your email address will not be published. Required fields are marked *

All comments posted become a part of the public domain, and users are responsible for their comments. This is a moderated site and your comments will be reviewed before they are posted. Read more about our comment policy »

Page last reviewed: October 28, 2021
Page last updated: October 28, 2021