5
it best to create a “calendar view” heatmap (Figure 1C)
which ran some simple calculations and displayed the aver-
age dust level for that day as one of three options: optimal,
caution, or alarm. The dust level for that day is generated
using the raw data from the sensor and averaged from 8
a.m. to 4 p.m., which is indicative of that mine’s normal
time of operation and a standard 8-hour shift. Working
closely with the mine, we began to discuss the key factors
and assumptions that need to be answered to simplify a
whole day’s data down into one number and its colored
risk classification (red for alarm, pink for caution, or white
for optimal), including: How many hours does any single
worker spend in the microenvironment around the sensor?
How many alarm levels are needed? Are there already estab-
lished regulatory limits? What will the actions be for an
“alarm” day? Once we had the necessary conversations and
we found, with the mine industrial hygienist, a comfort-
able status with the necessary assumptions, we began plac-
ing the calendar view heatmaps on the first page of the
weekly report for each sensor. If there was no red on the
weekly report, then they could simply move on to the next
sensor and know that the first sensor was good last week.
When a day gets flagged as red, then the mine would be
able to work backward to the hourly heatmap to try and
isolate the event to see if it was high levels for a long time
or a spike event that caused the day to be red. Then, when
Figure 1. Progression of data visualizations from line charts (A) to heatmaps (B), and finally
calendar view heatmaps (C)
it best to create a “calendar view” heatmap (Figure 1C)
which ran some simple calculations and displayed the aver-
age dust level for that day as one of three options: optimal,
caution, or alarm. The dust level for that day is generated
using the raw data from the sensor and averaged from 8
a.m. to 4 p.m., which is indicative of that mine’s normal
time of operation and a standard 8-hour shift. Working
closely with the mine, we began to discuss the key factors
and assumptions that need to be answered to simplify a
whole day’s data down into one number and its colored
risk classification (red for alarm, pink for caution, or white
for optimal), including: How many hours does any single
worker spend in the microenvironment around the sensor?
How many alarm levels are needed? Are there already estab-
lished regulatory limits? What will the actions be for an
“alarm” day? Once we had the necessary conversations and
we found, with the mine industrial hygienist, a comfort-
able status with the necessary assumptions, we began plac-
ing the calendar view heatmaps on the first page of the
weekly report for each sensor. If there was no red on the
weekly report, then they could simply move on to the next
sensor and know that the first sensor was good last week.
When a day gets flagged as red, then the mine would be
able to work backward to the hourly heatmap to try and
isolate the event to see if it was high levels for a long time
or a spike event that caused the day to be red. Then, when
Figure 1. Progression of data visualizations from line charts (A) to heatmaps (B), and finally
calendar view heatmaps (C)