Entry Name: HKUST-VastGp-MC2

VAST Challenge 2016
Mini-Challenge 2

Team Members:

Xinnan DU, Hong Kong University of Science and Technology, xduac@connect.ust.hk PRIMARY

Zheqing YU, Hong Kong University of Science and Technology, zyuaf@connect.ust.hk

Yuan CHEN, Hong Kong University of Science and Technology, ychencd@connect.ust.hk

Haoan FENG, Hong Kong University of Science and Technology, hfengac@connect.ust.hk

Ziyu WANG, Hong Kong University of Science and Technology, zwangbm@connect.ust.hk

Shaoyu CHEN, Hong Kong University of Science and Technology, schenan@connect.ust.hk

Chunyan BAI, Hong Kong University of Science and Technology, cbai@connect.ust.hk

Siwei Fu, Hong Kong University of Science and Technology, fusiwei339@gmail.com

Yuanzhe CHEN, Hong Kong University of Science and Technology, ychench@connect.ust.hk

Huamin QU, Hong Kong University of Science and Technology, huamin@cse.ust.hk

Student Team: Yes

Tools Used:

GAStechVis, developed by the UG student team led by HKUST visGroup for the challenge.

Approximately how many hours were spent working on this submission in total?

300 hours

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2016 is complete? Yes

Video

ihome.ust.hk/~zyuaf/vast.mp4

Questions

MC2.1 – What are the typical patterns visible in the prox card data? What does a typical day look like for GAStech employees?

In general, during weekdays, employees can be divided into three groups, i.e. those working during daytime (Group A), at night (Group B), and after midnight (Group C).

l Group A: They go to work between 7:15 to 8:00, have lunch break for an hour at noon and leave the company around 17:00.

l Group B: These employees belong to technical departments (Engineering, Facility and Information Technology). They usually go to work between 16:00 to 16:30 and leave the company before 24:00.

l Group C: Two staffs from Facility department arrive around 00:00 and work for system maintenance. Then, they leave the company around 7:00 and 7:40 respectively.

Figure 1-1

Also, the prox mobile data shows that the robot’s route is nearly fixed, and only active during 9-10am and 2-3pm on weekdays. Morning’s route and afternoon’s should be the same, but at 2pm we can see that technical employees (except cshipp) only appear in conference room and meeting room, those detected in their offices are not from technical departments. The animation is shown in the video.

Figure 1-2

A typical day of GAStech employees vary based on the department, which can be divided into three main divisions: Management staffs (Administration, Executive and HR), Technical staffs (Engineering, Facility, and Information Technology) and security.

l Management staffs all belong to the Group A but their schedules are more flexible. Compared with nearly all technical staffs who take a break in a specific time, most of the managers spend the time working at their offices and a few of them go to deli. Their leaving time is also flexible.

l Technical staffs usually arrive between 7:15-7:45, have a 10 minutes’ break around 9:00 and go to zone 7 for meeting or communicate around 10:00. Then at 10:30, they will have a one-hour meeting, facilities meet at zone1’s conference room while IT and Engineer meet at zone 6’s meeting room. Then, they go for lunch in floor 1’s deli around 12:00 and return around 1 hour later. In the afternoon, IT staffs and engineers go to zone 6 to have a one-hour meeting and facilities staff meet at zone 1 at the same time. Some people take a break for 10 minutes around 4pm. They leave the company around 5pm. See figure 1-3(green represent Engineer, pink stand for Facilities and Orange represent IT)

Figure 1-3

Here are more details:

IT: their schedules are highly regular.

Ø Day-shift: It is similar to the typical technical staff’s schedule.

Ø Evening-shift: It is similar to the Group B’s schedules. Specifically, everyone goes to zone one around 17:20 for 10 minutes. At 18:30, they have a one-hour meeting at zone six of the second floor. Around 20:00, they go to deli to have dinner or take a break for one hour. Then, another 1-hour meeting at zone6, at 11pm, they pass through zone 1 and go back to zone 7. A staff named Sven goes to floor3 once a day to report progress.

Engineering: Employees’ work is routine.

Ø Day-shift: They arrive at their offices (located in zone-2 or zone-6,7 and zone1) before 7:40. Engineers in zone 2 and zone 1 will go to zone 7 to have a meeting or communicate around 10am.

Ø Evening-shift: For employees working at night, they perform the same routines as IT despite their different working schedules. They have a half-hour meeting at 7:30pm, rest for an hour between 20:00 and 21:00, go to zone 6 at 22:00, leave zone 6 for zone 1 at 23:00 and then leave the company before 24:00.

Facility: (excluding the two who does nightshift)

Ø Day-shift: For most employees in Facilities department, their routines are almost the same as IT staffs except that their meeting place is different. Some employees will work on the third floor, who act as bridges connecting this department with Administration and Executive departments. Besides, some employees will only work on the first floor and leave the company between 14:00 and 15:00.

Ø Evening-shift: For employees working at night, some of them take responsibility to examine equipment on the three floors, leaving around 23:00 and some of them work in their offices, leaving around 24:00.

l Security: Employees can be divided into three groups. One group consists of two people, mainly responsible for the safety on the first floor. They will first go to floor2’s security office, maybe it is the general security office and they have to take work attendance there. They check all conditions every three hours. One takes responsibilities on the third floor, checking all conditions in four hours. The rest check zone 7 at 9:30 and 15:00 for around 10 minutes

See figure 1-4

Figure 1-4

MC2.2 – Describe up to ten of the most interesting patterns you observe in the building data. Describe what is notable about the pattern and explain its possible significance

1. The HVAC electric demand power is proportional to total electric demand power and total demand power mainly depends on the HVAC electric power because whenever the HVAC electric power changes, the total demand power also changes with the same pattern. The Correlation Matrix also shows that on every floor, VAV_SYS AIR LOOP INLET Mass Flow Rate shares the largest correlation with VAV_SYS SUPPLY FAN OUTLET Mass Flow Rate (floor1: 0.948, floor2: 0.998; floor 3: 0.995), which in turn verify our assumption using the visualisation method. (Figure 2-1)

Figure 2-1

2. VAV_SYS Air Loop Inlet Mass Flow Rate (Total flow rate of air returning to the HVAC system from all zones it serves, Figure 2-2) and VAV_SYS Supply Fan Outlet Mass Flow Rate (Total flow rate of air delivered by the HVAC system fan to the zones it serves, Figure 2-3) share similar patterns. Generally speaking, the patterns between these two data can be categorized into two theme: working and idle. In working pattern, outlet is always 0.3 kg/s larger than inlet mass flow rate and time spans from 6:00 a.m. to 22:00 p.m., corresponding to weekday's daytime. For the other pattern, idle, inlet mass flow rate is close to zero and outlet fluctuates intermittently. This pattern is energy-conservative. These patterns are significantly useful for anomalies detection and analysis.

Both VAV_SYS Air Loop Inlet Mass Flow Rate and VAV_SYS Supply Fan Outlet Mass Flow Rate held the feature that the values descended from floor 1 to floor 3.

3. In general case, Thermostat Cooling & Heating Setpoints is inversely related. However, when SUPPLY INLET Temperature is high，Thermostat Cooling & Heating Setpoints will increase and decrease in the same pace , and cooling and reheat coil power will increase correspondingly

Figure 2-4

4. Cooling coil and reheat coil influences each other. There exists some auto-thermal-balance system to keep the temperature stable. When the reheat coil power increases, the cooling power also increases to against the increment of temperature. Specially, the two highest impulses suggest the unusual changes in cooling and heating set points (Figure 2-5-1, 2-5-2).

5. Damper position, to some degree, is determined by the outdoor mass flow rate and influenced by the supply inlet mass flow rate. Therefore, damper position looks similar to mass flow. In each HVAC zone, the SUPPLY INLET Mass Flow Rate is highly related to VAV REHEAT Damper Position, with correlation close to one. The SUPPLY INLET Mass Flow Rate of F_3_Z_9 is abnormally higher than those in other zones and the REHEAT Damper Position of F_3_Z_9 form a different pattern compared the others. (Figure 2-6).

Figure 2-6

6. The Equipment power is stable and has strong correlation with time and employee activities. It is high from 5:30 to 22:30 on weekdays and remain low on weekend. Interestingly, on floor 3 the server room consumes significantly large amount of power comparing to others and the power consumption remains high at 30,000W all the time.

7. The Hazium concentration on the second floor is relatively high among these three floors while it is the lowest on the first floor. It is because most of the staffs and electrical equipment are located on the second floor. (Figure 2-7)

Figure 2-7

8. Magnitude relationships in general:

The Supply Inlet Mass Flow Rate behaved similarly in the majority of the zones on floor 3 except F_3_Z_9 (Server room), where the value was always much larger than those in other zones. In addition to this, VAV Reheat Damper Position and Supply Inlet Mass Flow Rate have a strong correlation (R > 0.99) which means that their trends are almost the same.

The maximum value of VAV_SYS Outdoor Air Mass Flow Rate over the whole period on floor 1 is the largest and irregular while that on floor 2 and floor 3 were almost the same.

VAV_SYS Air Loop Inlet Temperature is always larger than VAV_SYS Supply Fan Outlet Temperature, and their shared similar behavior on each floor.

Figure 2-8

9. Thermostat Heating Setpoint and Thermostat Cooling Setpoint:

In F_3_Z_9 (Server room), these two parameters are smaller than the average values of those in other zones. In addition, they are stable and rarely change (only changed from 6.7 to 6.9) during the fourteen days.

Generally, the tendencies of Thermostat Heating Setpoint and Thermostat Cooling Setpoint are opposite, except that on 6.7 and 6.8, they change in the same way.

Thermostat Cooling Setpoint is always larger than Thermostat Heating Setpoint in the same zone.

Figure 2-9

10. Thermostat Temp and Supply Inlet Temperature:

When Supply Inlet Temperature vibrates regularly, Thermostat Temp is kept stable and high, while when Supply Inlet Temperature keeps low, Thermostat Temp decreases and becomes low. However, from 6.7 to 6.8, the two building data have negative correlation which might result from the relationship between Thermostat Heating and Cooling Setpoint (both reach high values during that period).

Figure 2-10

MC2.3 – Describe up to ten notable anomalies or unusual events you see in the data. Describe when and where the event or anomaly occurs and describe why it is notable. If you have more than ten anomalies to report, prioritize those anomalies that are most likely to represent a danger or serious issue for building operation.

(1) Analysis on abnormal HVAC events on June 7th and June 8th

A lot of HVAC data goes abnormal on June 7th and June 8th. It seems that day and night are reversed in the air conditioning system. The parameters related to air conditioning reverse in that two days and that cause more anomalies.

For example:

(1) Supply inlet mass flow rate and VAV reheat damper position

Supply inlet mass flow rate and VAV reheat damper position should be relatively high during the day and damping after midnight. However, on June 7th and June 8th, the scale is extremely low

Figure 3-1

(2) Reheat coil power

The Reheat coil power reaches unexpected high value on 6.7 and 6.8, as labeled in the image below, which results in the fluctuation of the electric demand power. Similar anomaly happens more severely and lasted longer during 6.11 and 6.12

Figure 3-2

(3) Thermostat Temp

In the first week, temperature has natural and regular pattern. In other word, the value is high in the daytime while become lower at night within certain degree. However, anomalies occurr on 6.7 and 6.8. The difference between daytime temperature and night’s become large. In normal days, nights’ temperature is higher however on June 7 and 8, things are in the reverse way. After these two days, the regular patterns recover.

Figure 3-3

(4) Thermostat Heating Setpoint

Mostly behaves reversely with regard to Thermostat Temp and cooling setpoint. (i.e. when the former one is relatively small, the latter one is relatively large and vice versa). Except that on 6.7 and 6.8, all three of them behave similarly; In F_3_Z_8, the tendency on 6.9 is different from other zones on floor 3 but much more similar to that on floor 2.

Figure 3-4

The underlying causal relationship might be: SUPPLY INLET Temperature -> Thermostat Cooling & Heating Setpoints -> Reheat Damper Position & Supply Inlet Mass Flow Rate -> Reheat Coil Power & Thermostat Temp. The detailed illustration is as following:

(1) As Figure 3-5 shows, the Heating Setpoint in F1_Z1 becomes higher right after the Supply Inlet Temperature reaches the peak of 40 degree. Similar pattern can be observed on Figure 3-5

Figure 3-5

(2) Figure 3-6 shows that when the Heating Setpoint becomes higher than usual, position of the zone’s air supply box stays partially closed (VAV Reheat Damper Position) instead of switching on/off, which is the usual pattern. The Reheat Coil Power also acts in an unusual way correspondingly.

Figure 3-6

(2) Hazium Concentration

Hazium, a chemical gas with potential damage to human health, has abnormal concentration peak on 6.3, 6.7, 6.9, 6.13, as labeled in the image bellow. The unusual hazium concentration mainly happens in these four zones: F_1_Z_8A, F_2_Z_2, F_2_Z_4 and F_3_Z_1 (Figure 3-1). Our Neural Network classifies four Hazium Concentration into a same class alone, which means there isn’t any other building element strongly related to Hazium Concentration. Also, Hazium Concentration on the same floor has strong correlation (e.g. sensors on the 2nd floor have correlation of 0.994), while those across floor have less significant correlation (around 0.5).

Figure 3-7

(3) Difference HVAC behaviors between Floor1,2 and Floor3

Zones covering corridor in F1 and F2 has similar pattern in terms of SUPPLY INLET Mass Flow Rate and VAV REHEAT Damper Position. However, corridor in F3 has its own pattern of these two HVAC fields. This may indicate that F3 has its own HVAC system compared with that on floor 1 and 2 (Figure 3-8)

Figure 3-8

(4) Equipment power:

F_3_Z_2: periodic on weekdays and stayed the same on weekends except that there is a pulse on 6.11 during 8:30-11:30. F_3_Z_8: stays the same during the first weekend but there are two pulses on the second weekends (one of Saturday and the other on Sunday) (Figure 3-9)

Figure 3-9

(5) Return outlet CO2 concentration:

Abnormally large on 6.7 and 6.8. Besides, in F_1_Z_3, F_1_Z_4, F_1_Z_5 and F_1_Z_7: abnormally large on 6.5 and 6.6. CO2 concentration is supposed to be related with the population, as shown in Figure 3-10

Figure 3-10

(6) – (7) Proxout data:

Anomalies in the records of minority proxy card:
 

(6) June 6 – June 8:

On 6.6, Geneviere Florez (gflorez, Administration): stayed in F_1_Z_1 from 24:00 to the early morning. But it suddenly appears in F_1_Z_4 at 7:56 on 6.7.

On 6.7, at 23:59, Clemencia Whaley (cwhaley, Engineering) disappear in F_2_Z_4, Adan Morlun (amorlun, Facility) disappear in F_2_Z_4, Dante Coginian (dcoginian, Facility) disappear in F_2_Z_1. They appear suddenly on 6.8.

On 6.8, Clemencia Whaley (cwhaley, Engineering) disappear at 22:35. It is possible that there is something wrong in the elevator (might due to artificial issues), leading to the record mistake in and around it. And the same thing happens to him 6 times.

(7) During the second weekend:

On 6.10, Clemencia Whaley (cwhaley, Engineering) disappears at 22:50 in F_2_Z_1. On 6.11, Mat Bramar (mbramar, Administration) enters at 8:30 and directly goes into F_3_Z_6 and leaves at 11:30. Orhan Strum (ostrum, Executive) enters at 9:01 and goes into F_3_Z_3 and F_3_Z_6 and leaves at 13:30. On 6.12, Linda Lagos (llagos, Administration) enters at 8:00 and goes into F_3_Z_3 and F_3_Z_6 and leaves at 12:00. There are abnormal building data during the second weekend, and these people might be the cause.

MC2.4 –– Describe up to five observed relationships between the proximity card data and building data elements. If you find a causal relationship (for example, a building event or condition leading to personnel behavior changes or personnel activity leading to building operations changes), describe your discovered cause and effect, the evidence you found to support it, and your level of confidence in your assessment of the relationship.

1. Electric Power

The electric power consumed by each zone is consistent with the events happen in the building. The behavior can be illustrated by the following examples:

(1) When the flow of people was dense, the electric power tends to increase and vice versa.

On weekdays, in the morning, electric power increases (as equipment turned on) at 7:00 when the first person arrives. At around 18:45, the value becomes small because most of the employees have left the building; on weekends, there were few people, so the electric power was relatively low.

Confidence level: 80.60% of the data satisfies the hypothesis with a tolerance as 1000 (within 5% of the value of Total Electric demand power and HVAC Electric demand power).

(2) The following figure shows the change of temperature in Floor1_Zone1(HVAC zone), where the canteen of the building is located. It reaches the peak each day at around 12:30pm.

Figure 4-1-1

Meanwhile, the corresponding proximity zone, Floor1_Zone2, receives a large flow rate at the same time (perhaps lunch time for employees), which leads to the raise of temperature.

Figure 4-1-2

2. Hazium Concentration

The hazium concentration may have something to do with the human behavior. As the figure shows, there is a peak of hazium concentration on June 11st 18:00pm (Saturday). Also, there are only two people entering the company that day, Brama and Strum. There might be some human factors behind this.

Figure 4-2-1

Figure 4-2-2

3. Air flow

Floor 1 & 2: If the proximity data is intense, the value of outdoor air flow rate is stable. When there are fewer people in the proximity zone, the flow rate vibrates between zero and its default value. This may because that the HVAC system intends to save energy during the night.

Floor 3: When the value of floor 1 and 2 vibrates, the value on floor 3 keeps low.

Confidence level: 59.10% of the data satisfies the hypothesis. (In the test, few people flow: no more than 3 prox out records in a period of 5 minutes)

Figure 4-3

4. Floor Power

Floor power in corridors and areas with sparse flow of people tend to keep regular patterns all the time.

Figure 4-4

5. Temperature

On June 6th and June 7th 7:00-7:30, the reheat coil power was extremely high, which in turn results in the raise of temperature. We also observe that on June 7th, as figure5 shows, Florez enters the building at 0:00 and the record stays there until 7:00, when the regular office hour of Florez should start. This is quite abnormal. The HVAC data begins to go abnormal right after Florez goes to his own office, which could be suspicious.

On 6.7 and 6.8, despite the abnormities that most building data had, Supply Inlet Temperature has a sharp pea, indicating that it is likely to be the cause. The sensor which detects Supply Inlet Temperature gives out an extremely high output and influences the Thermostat Heating and Cooling Setpoint. Then Reheat Damper Position and Supply Inlet Mass Flow Rate change correspondingly, which finally leads to the abnormal behaviors of Reheat Coil Power and Thermostat Temp.

Figure 4-5