Fujitsu estimates workers' heat stress levels with new AI-based algorithm
Jul 12, 2017
Development of algorithm contributes to strengthened summer safety in the workplace
Tokyo, July 12, 2017: Fujitsu Limited today announced the development of a new algorithm to estimate ongoing heat stress in workers, such as security guards, using its artificial intelligence technologies, known as Human Centric AI Zinrai. Developed together with Fujitsu Laboratories Ltd., Fujitsu aims to enhance solutions deployed on its digital business platform, "Fujitsu Digital Business Platform MetaArc"(1) , that use IoT to support on-site safety management. This new algorithm will be available from the end of July. Also from June through September, Fujitsu is implementing solutions incorporating this new algorithm in a field trial with security guards at its own Kawasaki Plant.
Previous algorithms estimated heat stress levels(2) using a device worn on the arm to measure data such as humidity, temperature, as well as increases in pulse, and were primarily used in manufacturing and at construction sites. As the new algorithm can now estimate heat stress levels that accumulate over time, it is suitable for employees who continually work outdoors in the summer. Combined with the previous algorithm, it can be used to manage safety in a wide variety of situations.
In order to estimate the level of accumulated heat stress, and because a diverse range of data is handled in which correlations are difficult to discern, the new system uses AI technology to execute machine learning based on expert knowledge. In so doing, Fujitsu developed a logic in which AI can extract the characteristics of high heat stress to enable estimates such as those made by labor science experts.
In light of the knowledge gained from implementing this new technology, Fujitsu is accelerating safety management support for customers across a variety of industries. Fujitsu will help connect people, things and infrastructure across boundaries between companies and industries through "MetaArc" which will realize the customers' digital transformation, enhancing their competitiveness and creating new value.
In recent years, the number of workplace casualties(3) due to heat stroke in Japan has remained in the range of 400-500 people per year, even after 2010, which saw the worst heat wave on record. In 2016, the number of workplace casualties due to heat stroke reached about 450 people(4) , leading companies to take measures to prevent such incidents.
Going forward, the demand for heat stroke countermeasures is expected to increase rapidly for security and other operations at outdoor events and worksites in the summer.
Fujitsu developed this new AI-based algorithm to enhance safety protection for employees working in extreme environments.
Features of the New Algorithm
In addition to the existing data of humidity, temperature and pulse, the new algorithm estimates the level of heat stress based on new data, such the amount of activity, as well as data that shifts over time.Since machine learning is appropriate for making estimates from a wide variety of data with unclear correlations, Fujitsu developed a logic in which AI can extract the characteristics of high heat stress from the stock of actual data and data evaluated by experts. This has enabled the algorithm to estimate accumulated heat stress in the same way that labor science experts would, enabling users to observe the status of individual employees in situations that do not require a great deal of activity, such as security guards who must spend long hours standing in the hot sun.
Summary of the Internal Field Trial
Fujitsu has applied its solution incorporating the new algorithm to monitor those entering its Kawasaki Plant as well as security guards on site. By conducting this internal trial from June through September, months of extreme heat and humidity, Fujitsu is gaining the necessary knowledge to eventually contribute to the safety of employees at customer facilities.
Sites where security and other duties typically take place may be locations where workers are susceptible to heat stress. However, changes in physical condition vary according to the individual, making it difficult to take uniform measures. This newly developed algorithm makes it possible to estimate the accumulation of heat stress on a per person basis, to tailor ways to protect people based on individual conditions.
Image of the internal field trial
This solution will be exhibited at "Heat Solution Tokyo 2017," to be held at Tokyo Big Sight, July 19-21.
A digital business platform combining Fujitsu’s system engineering knowledge and experience with cutting-edge technology, including the cloud, mobile, analytics, IoT and AI.
[ Heat stress level
The Japanese Society of Biometeorology calculates heat stress levels from temperature, humidity and pulse, based on the relationship between WBGT (Wet-Bulb Globe Temperature), air temperature and humidity. The heat stress condition is estimated into four stages of safe, low, medium and high heat stress.
[ Number of casualties
Number of deaths and people who took four or more days off work due to job-related illness.
[ In 2016, there were approximately 450 casualties due to heat stroke in the workplace
"Occurrences of Worker Fatalities Due to Heat Stroke," June 2017 report from the Ministry of Health, Labour and Welfare in Japan. http://www.mhlw.go.jp/file/04-Houdouhappyou-11303000-Roudoukijunkyokuanzeneiseibu-Roudoueiseika/hasseijoukyou_1.pdf (Japanese)