Entry Name: “PKU-Yu-MC2”

VAST Challenge 2016
Mini-Challenge 2

Team Members:

Yu Zhang, Peking University, yuzhang94@pku.edu.cn     PRIMARY

TangZhi Ye, Peking University, yetangzhi66@gmail.com

Siming Chen, Peking University,  simingchen3@gmail.com

Chufan Lai, Peking University,  chufan.lai.1990@gmail.com

GuoZheng Li, Peking University, liguozhengsdu@gmail.com

Lu Feng, Peking University, 1200012795@pku.edu.cn

QiangQiang Liu, Peking University, lqqyeah@gmail.com

Shuai Chen, Peking University, seinchen@foxmail.com

Ren Zuo, Peking University, zuoren@pku.edu.cn

Zhuo Zhang, Qihoo 360, zhuangzhuo@360.cn

Zhanyi Wang, Qihoo 360, wangzhanyi@360.cn

Xin Huang, Qihoo 360, huangxin-xy@360.cn

Fengchao Xu, Qihoo 360, xufengchao@360.cn

Yu Li, Qihoo 360, liyu-safe@360.cn

Shunlong Zhang, Qihoo 360, zhangshunlong@360.cn

Qiusheng Li, Qihoo 360, liuqiusheng-s@360.cn

Xiaoru Yuan, Peking University, xiaoru.yuan@pku.edu.cn (Advisor)

Student Team:  NO

Tools Used:

EXCEL, MATLAB, and own developed tools.

Tableau is used for data exploration.

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

500 hours.

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

Video

http://vis.pku.edu.cn/pku-qihoo-mc2-video.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?

Limit your response to no more than 6 images and 500 words.

1. Three types of employees

There are three different types of employees according to their active time in the building. They can be divided morning shift, day shift, and night shift. Employees do not change their working time in the two weeks.

The majority of the employees (no less than 97 employees) take day shift. They come to the company from 7:00 to 8:00, and leave at about 17:00.

A few employees (no less than 15 employees) take night shift. They come to building at about 16:00 and leave at about 24:00.

Only two employees take morning shift. They are both of facilities department. They work from about 0:00 to about 7:00. They are only active on the 1^st floor.

2. Employees take break on the weekends

Most employees do not come the company at the weekends. Only 4 employees have records in the weekends. Three of them are administrators and another is executive.

3. Weekday daily meeting for day shift employees

The meeting is at around 10:45-11:30, 14:30-15:10, 18:45-19:15, 22:15-22:55, and the participant depends on the shift. At the meeting time, nearly all the staffs on the 2rd floor meet at two places (conference room and Mtg/Training). The location is different according to the department category (information technology and engineering meet at #2700, facilities meet at #2365).

As for the nightshift meeting, the categories of night shift staff are mainly engineering, information technology, and facilities. There is always a small meeting for the nightshift before the work at around 18:45, and lasts around 30 minutes. Still, information technology meet at #2700 and facilities meet at #2365.

4. From 12:00 to 13:00 and around 20:00, the staff eat lunch on floor1

The meal schedule of the engineers are regular. Besides, the lunch time of Info technology, Engineers, and Facilities are more regular than the administrator, whose lunch time are also a little earlier than the others. The executives hardly come to deli for their lunch.

5. Judging from the spatiotemporal distribution of the prox card detected by the robot, the movement of the robot can be recovered. The robot moves on a fixed route every workday twice, from 9:00 to 10:00 and from 14:00 to 15:00.

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 what you can about the significance of the pattern.

Limit your response to no more than 10 images and 1000 words.

1. The working pattern of HVAC system:

1) For floor 1 and floor 2, 6:00 to 22:00 on workday, the HVAC systems of all three floors work continuously.

2) For floor 1 and floor 2, 0:00 to 6:00 and 22:00 to 24:00 from Monday to Friday, the system stops periodically. After working 1 hours, it would rest for 5 minutes. The systems of floor 1 and floor 2 work on weekend.

3) For floor 3, the system doesn’t work on weekend.

At night and in the weekends, the Availability Manager Night Cycle Control Status attribute of the floor denote whether the HVAC system is working (2 for open, 0 for close). From this attribute, we can find the cause of the pattern of resting 5 minutes an hour when the HVAC system is running when it is not working hours for the majority of the employees (22:15 to 6:05).

Attributes of the HVAC system has different patterns on weekdays and on weekends because of the pattern of control status.

This general working pattern of HVAC system is significant in that it explains why many attributes have sudden drops periodically, such as HVAC zone power.

2. Relationships among parameters

1) Thermostat heating and cooling setpoint tunes the thermostat temperature in the zone

During the period mentioned above when HVAC system works, if thermostat temperature of an HVAC zone is higher than cooling setpoint, the thermostat temperature would become lower as the HVAC system works; likewise for heating setpoint, when the system works, the temperature would go higher if it is below heating setpoint.

The HVAC system indirectly controls the thermostat temperature by changing the supply inlet temperature and supply inlet mass flow rate. In this way, the system can control the thermostat temperature of air in the room.

When Thermostat temperature is lower than heating set point, the REHEAT COIL would start to work (reflect on REHEAT COIL Power) to tune the temperature of inlet air (reflect on supply inlet temperature).

Supply inlet temperature is controlled by reheat coil power and supply fan outlet temperature of the floor. Supply fan outlet temperature is the air supply temperature of HVAC system. When the HVAC system is working normally, the supply inlet temperature is between 12-14 degree and when HVAC system is not working, its temperature is near Drybulb Temperature.

2) The cooling coil power is positively related to Drybulb temperature

When the temperature is high, the cooling power has to be high to reduce the temperature of outdoor air to maintain a stable temperature. Besides, cooling coil power is related to supply fan outlet mass flow rate. To supply more air to the HVAC zones on a certain floor, the floor’s cooling coil has to increase.

3) Inlet Mass flow Rate and Inlet Temperature effect indoor temperature together

4) Outlet temperature maintains at about 12.8 degrees during normal work of HVAC，while equals to outdoor temperature (Drybulb temperature) when HVAC is not working.

5) Outdoor Air Mass Flow Rate is equal to the difference between Supply Fan Outlet Mass Flow Rate and Air Loop INLET Mass Flow rate. And the difference will affect outdoor Air flow Fraction.

6) Normally, when the HVAC system of the floor is working, the Air Loop Inlet Mass Flow Rate is high. Notice that an anomaly happened on 6.7 from 0:00 to 8:00, and from 22:00 to 8:00 on 6.8 morning. At that time, the Air Loop Inlet Mass Flow Rate has many pulses when the system was in non-working hours. This pattern happened on all the three floors at that time.

The physical and statistical relationships among attributes helps to find to the causality of events.

3. Hazium pattern

The 4 Hazium sensors have many similarities in the temporal distribution of their peaks. For example, many of them have peaks near 6.3 10:00, 6.7 15:00, 6.11 15:00. Since high Hazium concentration can be viewed as an important anomaly, the pattern of Hazium concentration is an important entrance of exploring the data.

4. Lighting pattern

On floor1, zone8A (corridor), zone8B (corridor), and zone3 (main entrance), the light is constantly on, and therefore the light power is constant. On floor2, zone12A (corridor), zone12B (corridor), zone12C (corridor), the light is constantly on. On floor3, light in zone12 is never on.

5. Cooling pattern

The heating power of the building is always 0, which means that the building is continuously cooling. The cooling coil power, fan power, inlet mass flow rate have high correlations, which implies that in the two weeks, sending wind is aimed at cooling.

6. Outdoor air related attribute ranking

The outdoor mass flow rate of floor1 is generally higher than floor2 and floor3. This might due to the main entrance opening to the outside, which adds to the ventilation of floor1. Generally speaking, the Outdoor Mass Flow Rate ranking is floor1 > floor2 > floor3, while the Outdoor Air Flow Fraction ranking is floor1 > floor2 = floor3.

Besides, the Outdoor Air Mass Flow Rate and Outdoor Air Flow Fraction have high positive correlation.

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.

Limit your response to no more than 10 images and 1000 words.

1. Abnormal temperature related HVAC attribute reading in floor3-zone8

On 6.9 at 10:10, many attributes of floor3 zone8 that are related to temperature suddenly become abnormal, and this anomaly continued to 22:45. The heating and cooling setpoint was set low, but oddly enough, the heating power, air temperature, and supply inlet temperature burst high.

Notice that this event happened slightly after Hazium reading of floor3-zone1 reached a peak on 6.9 at 8:00.

What’s more, the heating setpoint and cooling setpoint of floor3 zone8 departed from the majority of other zones in floor3 at 10:00.

2. Rare meeting event

Forluniau Carla (cforluniau001) and Frost Jeanetts (jfrost001) in department of Administration, together with Mies Haber Ruscella (rmieshshaber001), went to floor3 zone2 at around 10:00 on 6.1, and left at around 12:00.

This was the only time these three persons went to floor3 in the 14 days.

And during their stay in floor3 zone2, the Inlet Air Damper Position decreased, and the damper position returned to normal pattern after they left.

3. Anomaly in self-connection of the same zone

As is displayed in the connection matrix, floor1-zone1 (f1z1), zone3 (f1z3) and zone7 (f1z7) are connected with themselves, which is abnormal. The degree of f1z1’s self-connection, is equal to the degree of edge connecting f1z1 and f1z3, and judging from the trajectory, the 30 persons moving between f1z1 and f1z3, and the 30 persons detected moving from f1z1 to f1z1, are the same group of people. Besides, it can be deduced that f1z1, as a big zone containing the entry of building, have two prox card detectors.

What’s more, the prox cards entering f2z7 (3163) and exiting f2z7 (3160) is not equal. For f2z7, an anomaly exist in Herrero’s (kherrero001) trajectory on 6-1: f1z1→ f1z1 →f1z4→f2z4→f2z1→f2z7→f2z3→f1z4→f2z4→f2z1→f2z7→f2z3→f2z7→f2z7→f2z3→f2z3→f2z7→f2z1→f2z7→f2z1→f2z4→f1z4→f1z1→f2z4→f1z4

4. Lost card used

After Young Patrick (pyoung001) lost his prox card and applied for a new card, his old prox card was still used, which is very suspicious.

Especially, the lost prox card pyoung001 was detected in 2345, which is the office of Bennett Loretta. It seems that Bennett Loretta took this card.

5. Abnormal server room visited

After further checking Bennett’s visiting sequence, we found he visited server room several times (6.2 - morning 8:00, 6.3, 15:00 for a short time), and there might be correlations.

6. Florez lost his card several times

From 5-31 – 6-3, Florez lost his card nearly one time one day. The number already recorded to gflorez005. Further, in 6-7, it seems that he (at least his card) didn’t leave the company on the night of 6-6.

7. Abnormal CO2 concentration on floor3

In 6-8 afternoon, the Return Outlet CO2 is abnormally high in many zones on floor3.

8. Absence of several people in the normal place he usually should be

For example, the robot’s detection and prox card information was conflicted in 6-9 for Vico (avico001). Barranco (ibarranco001) didn’t show up in the office on 6-10.

9. Administration members’ meeting

On 6-1, the administration department had a meeting, which started around 10:30 am, and ended around 12:30.

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.

Limit your response to no more than 10 images and 1000 words.

1. High Total electricity demand power in the normal work time (Confident)

At the normal work time, the parameters of the HVAC system is perhaps manually or automatically changed from the night cycle, and the electricity system is more active, and therefore the power used is higher.

2. Morning shift employees enter conference room leading to high equipment power (Confident)

Morning shift employees enter the conference room on floor 1 every workday at 3 and 6 o’clock, and caused the equipment power of this zone went high. It seems that every day they check the equipment in the conference room.

3. High HVAC attributes in deli at noon (Confident)

At noon many HVAC attributes of deli is high, because there are many employees staying in deli, and therefore the power used goes high.

4. Abnormal low temperature caused by gfloerz (Not so confident)

At 0:05:00 on 6.7 the cooling setpoint and heating setpoint of the majority of HVAC zones was significantly lower than normal value, which led to the sudden increase of the Cooling Coil Power of all the 3 floors. The persons that might lead to this event are shown in the gantt chart on the right.

Notice that only gfloerz005 stayed in the building at 0:05:00 and is in floor1 prox zone1, which has a room for server. We guess that the server can be used to tune the parameters of the HVAC system, and it was gfloerz005 that was tuning the parameters and caused the anomaly in this event.

In the following image, we can find that oddly enough, gflorez lost his card for 3 times in the first 3 days. And the night from 6.6 to 6.7 is the only time he stayed in the building at midnight

Setting the cooling setpoint (at around 15 degrees Celsius) and heating setpoint (at around 12 degrees Celsius) low led to the gradual decrease of the air temperature of many zones. Such situation sustained until around 7:05 on 6.7,when the cooling setpoint and heating setpoint was tuned to significantly higher than normal value, and therefore the Reheat Coil Power of many zones increased dramatically and the Supply Inlet Temperature of many zones also increased. The Inlet Mass Flow Rate and Damper Position drops meanwhile. The HVAC electricity demand of the building also burst up to provide energy for the sudden heating behavior, which lead to a sudden increase of the building’s total electricity demand.

At around 7:05, only a few employees (vawelon001, wvasco001, acampo001, lorosco001, gflorez005) were in the building, and therefore it might be that some of them tuned up the cooling and heating setpoint.

After the setpoint was tuned up, the Thermostat Temperature of many zones increased. And the Hazium sensor reading of all the 4 zones abnormally increased since 3:25 on 6.7. Slightly after this anomaly in temperature, on June the 7^th 14:55 the readings of four Hazium sensors reach a peak nearly simultaneously. It seem that this rise in Hazium concentration is related to abrupt change in temperature.