Daniel Alcaide, KU Leuven, daniel.alcaide@kuleuven.be, PRIMARY
Jansi Thiyagarajan, KU Leuven, jansi.thiyagarajan@kuleuven.be
Houda Lamqaddam, KU Leuven, houda.lamqaddam@kuleuven.be
Jaume Nualart, KU Leuven, jaume.nualart@kuleuven.be
Jan Aerts, KU Leuven, jan.aerts@kuleuven.be
Student Team: YES
Approximately how many hours were spent working on this submission in total?
320 hours
May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2016 is complete? Yes
Video
http://homes.esat.kuleuven.be/~dalcaide/VASTchallenge2016/video/Kuleuven-Alcaide-MC2.wmv
Alternative mirror: https://youtu.be/7YMuKo3vJ_w
Important notes:
MC2.1 – What are the typical patterns visible in the prox card data? What does a typical day look like for GAStech employees?
MC2.1-1. Shifts:
For most of the employees, working hours are from 9am to 6pm (see Figure 1.1). There are 2 additional shifts:
- 23h - 7h
- 7h - 17h
- 15h - 23h
The departments that have 2nd shifts are IT, Facilities, Engineering.
Most employees have 8h shifts. Some people (mostly executives) work half days (4 hours).
MC2.1-2. Office distribution:
Based on Figure 1.1, we can imply the following location of spaces across the floors:
Floor 1:
Floor 2:
Floor 3:
Figure 1.1: Employee presence (line) by department (vertical blocks at right) and floor (colour) based on the proximity data
MC2.1-3. Lunch breaks:
For most employees that work 1st shift, lunch hours are between 12 and 2PM. During that time, the majority of them will go to the office Deli located at floor 1. Specifically, the engineering, IT and facilities departments will gather there at around 1PM.
Workers from other departments have slightly more distributed eating times. Administration staff for example will start coming in the deli at around 11, and typically come in in small groups of one or two people.
Some employees working second shift - namely IT, facilities and engineering, eat during the evening. They typically go to the deli between 8 and 9PM.
Figure 1.2: Deli occupancy for three consecutive days per department
Most security workers are seen around the main entrance for the duration of the lunch hours. Knowing their last recorded data is in the entrance hall, we can assume that these workers leave the offices for lunch break.
Figure 1.3: Entrance hall occupancy for three consecutive days per department
The stair and lift usage patterns confirm these patterns: there are usage peaks every day around 9AM, 12PM, 2PM and 6PM.
Figure 1.4: Occupancy graph for each proximity zone during the 14 days
MC2.1-4. Work habits:
The distribution of employees across the different days of the week seems to follow regular patterns. However, employees from different departments do have different work habits.
The engineering employees seems to have daily meetings with IT members before lunch as well as in the early afternoon (zone b on figure 1.5).
Employees of the security department mostly work from their dedicated office. They perform what seems like rounds within the 2nd floor every 1-2 hours, during which their office is empty.
There are also security officers dedicated to the 1st and 3rd floor, where they work all day.
There is more or less interdepartmental mixing in different floors and proxy zones. The executive employees, for example, are rarely seen outside of their offices in the 3rd floor. Other positions like security, facilities, and administration can be seen throughout the building.
The security office (zone a on figure 1.5) is the most guarded one, as it is typically only entered by members of the security department. Offices 7 and 8 on the 1st floor are also reserved for security.
The executive employees do not all have an 8h shift. Many of them only come during the morning, which leads to the third floor typically having a higher occupancy in morning hours. This is visible on figure 1.6.
Figure 1.5: Occupancy graph per department for each proximity zone during the 2nd of July
Figure 1.6: Executive employees presence by floor during the tracked period
MC2.2 – Describe up to ten of the most interesting patterns you observe in the building data.
MC2.2-1. VAV Availability Manager Night Cycle Control and its impact
Figure 2.1.1: VAV Availability Manager Night Cycle Control Pattern
Figure 2.1.2: Impact of Night Cycle Value on Outdoor Air Flow (left: across all floors; right: per floor)
Figure 2.1.3: Impact of Night Cycle Value on Reheat Damper Position
MC2.2-2. Unique and Strange Reheating Pattern for Zone 1 in Floor 3 or the CEO office
Figure 2.2.1: P-F3Z1 Thermostat Setpoint
Figure 2.2.2: Overall Reheat Coil Power Pattern and P-F3Z1 specific Pattern
MC2.2-3. Water heater gas rate is generally high between 12:00 and 13:00
The water heater burns more gas between 12pm and 1pm. At the same time, the opposite is true for supply side inlet temperature, as the temperature of water entering the heater is generally low around that time. This pattern is consistent throughout except on Sunday during which there is no heating operation. Unfortunately, we don't have data on where the hot water is consumed. We assume this will mainly be in the Deli on the 1st floor.
Figure 2.3.1: Water heater gas rate and supply side inlet temperature
MC2.2-4. Power Consumption
We identified several patterns concerning power consumption in the building:
Power consumption for lights power and equipment align in most cases.
Figure 2.4.1: Lights Power Pattern for Floor 1 excluding corridors
MC2.2-5. Thermostat Setpoint
The value was set to 29.70°C for cooling which is normally expected to be around 24°C and for heating the value was set to 26.67°C rather than normal value of 21°C. We think this triggers the opposite function as one would expect the zone to be cooled when thermostat temperature indicates 26°C. The impact of this setpoint is reflected in the activation of the reheat coil.
Figure 2.5.1: Thermostat setpoint pattern (A: cooling; B: heating)
MC2.2-6. Exhaust Fans
Figure 2.6.1: Bath Exhaust Fan Power Pattern
MC2.2-7. Return Outlet CO2 concentration
Figure 2.7.1: Return Outlet CO2 concentration
MC2.3 – Describe up to ten notable anomalies or unusual events you see in the data.
Note: We calculated a "reference" value for every individual and sensor. Two references were generated for the individuals: one reflecting the most common order of zones that a given individual enters; the second reflecting the average length in each of these zones. For the sensors, the reference indicates the most common pattern across a day. To identify anomalies for a person/sensor on a particular day, we calculated the difference of that day with the reference.
MC2.3-1. The HVAC system seems to be not working properly on June 7th and 8th. The problem is generalized in all zones, except zones P-F3Z1 and P-F3Z12. When checking P-F2Z12 (Figure 3.1) we observe that:
Figure 3.1. Screenshot of the interactive interface for detecting abnormalities in building data on floor 2 zone 14 on June 7th
MC2.3-2. We detected the same fixed value for the heating and cooling thermostat during the whole weekend. The actual events were:
The HVAC system was switched on unnecessarily to keep the temperature around the value established by the setpoint.
Figure 3.2. Interface for detecting abnormalities in building data on floor 3 zone 3 on June 5th
MC2.3-3. Considering Hazium, anomalies include (but are not limited to; sorted by maximum ppm; see Figure 3.3):
Other increases of hazium levels (under 5ppm) can be seen in the Figure 3.3. According to specifications, the hazium concentration did not reach risk levels.
Figure 3.3. Interface for detecting abnormalities in Hazium data on floor 3 zone 1 on June 11th
MC2.3-4. We detected a strange and recurrent pattern in the CEO's office (Floor 3 Zone 1) from June 2nd to 13th:
This pattern is remarkable because the temperature is very extreme between 12am to 5am and 1pm to 12am when no one is in the office. This might indicate the inability of the CEO to master his thermostat, or a security breach if these setpoints are managed centrally.
Figure 3.4. Interface for detecting abnormalities in building data on floor 3 zone 1 on June 13th
MC2.3-5. The offices located in E-F3Z8 usually have a temperature around 24°C. On June 9th the thermostat set point was increased to 32°C from 10am until 11pm (Figure 3.5). This fact was only detected in this location so any employee could have caused this abnormality.
Figure 3.5. Interface for detecting abnormalities in building data on floor 3 zone 8 on June 9th
MC2.3-6. We detected that on the third floor’s corridor (E-F3Z11A, E-F3Z11B and E-F3Z11C) the equipment and light power were out of order during the 14 days. This may indicate that the facilities are new and the building is not completely ready (Figure 3.6).
Figure 3.6. Interface for detecting abnormalities in building data on floor 3 zone 11 on June 1st
MC2.3-7. Four employees were present in the building on the weekend. Three of them (mbramar001, lcarra001, and llagos001) work in Administration in P-F3Z3; ostrum001 is an Executive. We did not expect the presence of any employees during the weekend. Therefore, we categorize this event as an abnormality (Figure 3.7).
Figure 3.7. Interface for detecting abnormalities in proximity indicating the individuals in the building during the weekend
MC2.3-8. Some employees lost their proximity card during these 14 days. The abnormal case we consider here is user Geneviere Florez (gflorez001- gflorez005) who asked for a replacement card five consecutive times.
Patrick Young asked for a replacement card on June 2nd (pyoung002) while the previous card (pyoung001) was still used. This could be a security issue because the person who had possession of the first card had access to office 2345 and the server room (Figure 3.8).
Figure 3.8. Interface for detecting abnormalities in proximity selecting the users pyoung001 and pyoung002
MC2.3-9. The employees with card IDs gflorez005 and jsanjorge001 showed abnormal behavior. They were in the office at 12am (jsanjorge001 on June 2; gflorez005 on June 7) instead of their usual 7am. It looks like the cards were activated automatically in zone 1 but the cause of this event is unknown (Figure 3.9 shown the example for the user jsanjorge001).
Figure 3.9. Interface for detecting abnormalities in proximity selecting the user jsanjorge001
MC2.3-10. Two abnormalities were detected in the toilet of third floor, room 3998. The most likely explanation for this is that the proximity cards could have been forgotten in the toilet during several hours.
Figure 3.10. Interface for exploring the sequence of employees in proximity data selecting the user rpantanal001
MC2.4 –– Describe up to five observed relationships between the proximity card data and building data elements.
MC2.4-1. Due to the low number of employees in the offices during the weekend, their presence is recognizable in the building data. The examples found are listed below and Figure 4.1 and Figure 4.2 are used to illustrate the first case.
Figure 4.1. Interface for detecting abnormalities in proximity selecting the user lcarrara001
Figure 4.2. Interface for detecting abnormalities in building data on floor 3 zone 10 on June 5th
MC2.4-2. Concerning the Hazium concentration, user mbramar001, from Administration, was detected by the proximity sensors in P-F3Z6 on Saturday, June 11 between 08:32 to 11:30 (Figure 4.3). Her office has number 3100 located in that zone. The maximum concentration of Hazium was recorded on the same day. The first ppm appeared around 11:00 am. This indicates that she might be the person who put the Hazium next to the ventilation system. We consider two options: either in her own office, 3100; or the CEO's office 3000 because this one is adjacent and has a sensor that measures this chemical. This theory could make sense because the sensor in E-F3Z1 has the most prolonged concentration during the following hours out of the four Hazium sensors around the building (Figure 4.4) and user mbramar001 did not appear in the records of proximity data on the next Monday. We wonder if this is related to the fact that the thermostat is set to extreme values in the CEO's office: is it his/her secretary doing this?
Figure 4.3. Interface for detecting abnormalities in proximity selecting the user mbramar001
Figure 4.4. Interface for detecting abnormalities in Hazium data on floor 3 zone 1 on June 11th
MC2.4-3. Meeting rooms are only used when the employees have a meeting. We can relate proximity and building data showing who is attending the meetings and the energy consumption of the room.
The user mbramar001 was detected in the office ‘Lg Meeting’ (office 1050), located in P-F1Z5 (E-F1Z2) from 11:28am to 12:25pm on May 31 (Figure 4.5). The consumption of equipment and lights power increased during this period (Figure 4.6). We only detected one employee in that location. Therefore, if there were any other employees present at the same time, they did not have their cards with them.
Figure 4.5. Interface for exploring the sequence of employees in proximity data selecting the user mbramar001 on May 31st
Figure 4.6. Interface for detecting abnormalities in building data on floor 1 zone 2 on May 1st
MC2.4-4. We can relate the usage of the Conference room (office 1030) between building and proximity data:
Figure 4.7. Interface for exploring the sequence of employees in proximity data selecting the zone 6 in first floor on June 3rd
Figure 4.8. Interface for detecting abnormalities in building data on floor 1 zone 6 on June 3rd
MC2.4-5. There is only one shift working on the third floor. It is easy to recognize this pattern by using both proximity data and building data. Using proximity data, we can see the time when the employees are detected in the building, i.e. between 7:00am to 6:00pm (Figure 4.9 for P-F3Z1). Power consumption used by equipment and lights is also high during this period. Figure 4.10 shows the data for zone E-F3Z6 as a representative example of the floor.
Figure 4.9. Interface for exploring the sequence of employees in proximity data selecting the zone 1 in third floor on May 31st
Figure 4.10. Interface for detecting abnormalities in building data on floor 3 zone 6 on May 31st
