CSU-Wang-MC2

VAST Challenge 2016
Mini-Challenge 2

Team Members:

Junwei Wang, Central South University, 2419695965@qq.com

Yuan Zeng, Central South University, 844755848@qq.com

Juncai Li, Central South University, dreair@csu.edu.cn

Lin Li, Central South University, 125910192@qq.com

Qi He, Central South University, 978632333@qq.com

Xiangqian Wang, Central South University, 1546777063@qq.com

Ying Zhao, Central South University, zhaoying511@gmail.com SUPERVISOR

Fangfang Zhou, Central South University, zhouffang@gmail.com SUPERVISOR

Student Team：YES

 

Tools Used:

Python

D3.js

Echarts.js

Flot.js

MySQL

 

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

About 150 hours

 

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

 

Video:

VAST Challenge 2016 MC2 Answer Sheet.fld\CSU-Wang-MC2-VideoDemo.wmv

 

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

 

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.

 

There are 7 patterns definitively discovered through our 3 customize visualization tools as following.

1)     Work Shifts, when to work, there were morning shift, middle shift and night shift;

2)     Daily Routines, a staff when and where to do what;

3)     Group Activities, like attending meetings, rotating to inspect, gathering or scattering crowds etc.;

4)     Trajectories, including staffs and robot;

5)     Card Statuses, the card was kept or lost;

6)     Staff Communications, when and where the staffs met and stayed together for how long;

7)     Special Behaviors, the actions of regular meeting or being absent or other abnormal actions.

Next, we are respective to introduce the 3 tools and illustrate the 7 patterns in our tools.

 

1)     The TimePlaceChart(Fig 1.1) can present the work shifts , daily routines, and group activities, by personal or department.

the vertical axis stands for where an employee stayed for some while, divided into his own office, meeting rooms, deli, other’s office, and other locations he stayed. The horizontal is a time sequence. So a polyline makes an employee’s trace. The employee moves to a new place while the polyline extends and the span of polyline is the duration of the previous action. We can observe several workers’ location changing in the TimePlaceChart using legends. To special, if there are many curves overlap, it shows group activity.

 

Fig 1.1      The TimePlaceChart, select all legends to find department patterns(left), select one or more to find personal patterns(right)

 

2)     The Prox Map(Fig 1.2) is able to depict trajectories of staffs and robot in the building over the time.

The walking trajectory of the robot repeats in two days. Using colors to mark the status of prox card, and provides an interactive way to explore the communications among employees. We count the number of staffs for each zone in every 5 minutes, the same slot of the building data, to point out when and where people gathered and scattered. Fig 1.2 just demonstrates the symbols and glyphs of patterns, not representing the actual situation.

 

Fig 1.2      The Prox-Map

 

3)     The RadarView(Fig 1.3) is aimed to illustrate the time distribution of a staff staying in each area.

The left radar is rendered in a day and the right is in a week. When discerning the significant distinct between the left and the right, we would be motivated to analyze the people further in the day. We find special behaviors, like regular meetings of week, being absent, etc.

 

Fig 1.3      The RadarView

 

Those are just some representative views for exploring the prox card data in our visualization system. Additionally, three views can be cooperated by the time, the location, and the proxid. So there is a remarkable convenience that enables one to explore the data from diverse aspects at the same time.

 

Totally speaking, we find each department has its own schedule, while employees share the similar behaviors in the same department and the same work shift. A typical day of each employee from departments can be summarized as follows:

 

 

 

 

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.

 

Building data is the time series data sampled with the identical interval. Each property of energy zone is a continuous curve over the time. The trends of curves may indicate some kinds of events make the properties changed. We found that there are diversities of curves on different floors, while similarities on the same floor. So we analyzed the building data from floors to zones hierarchically. Then we focus on observing 15 properties related with the zone and the floor, including power parameters, temperature parameters, mass flow rates, CO2 concentration and damper position.

 

Through a considerable number of experiments, we summarized two methods can excavate patterns among properties. One is observing curve’s periodicity to discern the value of polynomial arithmetic, the other is simulating the differential action to discover the relationships between  properties during their trends fluctuation. Periodicity and trends of the time curve is non-linear differential attributes, so we make no sense when using linear methods to process building data. Accordingly, we designed the PropertyChart shown in Fig 2.1 to achieve above two ways to find nonlinear patterns of the building data. Plus, we designed another view in Fig 2.2 to combine each zone with its properties. With operations of normalization and transition. Fig 2.2 can interactively illustrates multiple curves to gain insight of properties’ evolution over the time. Besides, we alse can uncover some interesting latent patterns in Fig 2.2 and get them validated in the Fig 2.1

 

Fig 2.1            the PropertyChart highlight the cycles and trends to find patterns

 

 

Fig 2.2     Click the zone in maps and display all property curves of the zone.

 

1)     We discover equality relationship among properties, which only occurs in polynomial arithmetic with the same physical dimension.

We can recognize equality relationship in the PropertyChart through discerning the consistency of property curves in polynomial arithmetic. Benefiting from the distinction of property curves and the detailed description of data, we find out all useful equalities in building data in Fig 2.3.

 

Fig 2.3      The equality patterns we found

 

 

2)     We find out the interesting properties by the periodicityof curves and give an explanation via deliberately understanding the exact meaning of observed curves.

There is a description that how we found the supply power patterns of corridors, ServerRoom and elevators in Fig 2.4:

 

Fig 2.4      We find power patterns of corridors,ServerRoom and elevators.

 

3)     We discover the relationships among properties with different physical dimensions.

Properties with the same physical dimension may make useful statistic, and with the different may present hidden patterns by the periodicity. We find the curve of difference between Supply Side Inlet Temperature and Drybulb Temperature is quite similar with the Water Heater Gas Rate in Fig 2.5. We think it may be a pattern that the temperature of water before entering conditioning system is the same as the outside, and it get energy from the Water Heater Gas to reach what can be used in the system.

 

Fig 2.5      We guess the patterns of water heating

 

In general, the periodicity and trend are considered as the most important aspects of the curve similarity. Curves with similar ups and downs are similar. The periodicity is easy to be distinguished according to the corresponding wave, but the trend can’t be readily recognized, because it’s influenced by the numerical value of the fluctuation and different orders of magnitude from different physical dimensions. Using the normalization and sharpening operations in the PropertyChart(Fig 2.2) can detect whether curve is stable, rising, or falling with different undulate fitness.

 

4)     We find the way that the HVAC system adjusts the temperature and freshes air by comparing the trends of properties of the HVAC system in Fig 2.6.

When rising,  the CO2 concentration of one zone may become a factor of the high temperature and dirty air. The HVAC system turns the damper up to raise the ratio and flow rate of outside air.

 

Fig 2.6      We find how HVAC system deal with the air from the properties’ trends

 

5)     We find HVAC system’s working mechanism in Fig 2.7.

It shows that every floor installs one air conditioner, and each zone was equipped with independent terminal to adjust temperature. The cold air produced by the central air conditioner is able to be distributed to each zone according to its requisite-amount.

Fig 2.7 shows that there was only a fan power on a floor, but each corresponding zone had air velocity and their curve trends are analogous to the fan power. Further more, the similar relationships between the floor and the zones are also reflected in the air temperature and velocity. Therefore, we make the above speculation.

 

Fig 2.7      We find system’s working mechanism by comparing the properties of the zone and floor it belongs to.

 

6)     We detect the pattern of HVAC cooling power in Fig 2.8. We find power used by the HVAC system cooling coil is related to the product of indoor temperature difference and flow rate of the air HVAC system supply.

In Fig 2.8, The polynomial is related to zones, we found it based the previous polynomial, according to the HVAC work structure we inferred, the sum of the zone’s properties on the floor would be the parameter about the floor. And the trend is consistent with the cooling power that floor.

 

Fig 2.8      Combine the basic statistics to the complex one and expand the relationships

 

7)     In the same way, through the similarity of the property curves, we find:

a) Light power is related to the equipment power. It says in a room, parts of the equipment and lights the same switch. Turn on the lights can open equipment at the meanwhile.

b)Deli fan power is related to the night-cycle-control-status of floor 1. They share the same trend at night. Maybe it is to facilitate, at night Deli fans can be controlled through the control status of floor 1.

c) The cooling power is related with lights power, equipment power, CO2 concentration and drybulb temperature. It must be noted that, maybe it’s related to many factors, so the trend is not obvious in the descent phases and we find they are consistent during the rising phases.

 

8)     We discover the HVAC system can maintain the thermostat temperature by regulating the reheat coil power.

On the left of Fig 2.9, the thermostat temperature changes very little and the reheat coil power appear during the time. We suppose the air conditioner supplys cool air according to the cooling setpoints, and then the reheat coil works to heat the air in order to retain the constant temperature.

 

9)     We also detect another pattern that the HVAC system maintains the thermostat temperature by switching on and off the air conditioner.

According to the right Fig 2.9, when the room temperature decreased to a certain temperature, the damper would be closed automatically, the flow rate disappeared simultaneously, which means the conditioner had be turned off. The air conditioner would return to work and supply the cool air again until the temperature reaching the certain value.

 

Fig 2.9      Two different ways we found the air conditioner keep the thermostat temperature

 

10)  We infer that the Hazium may emerge when the CO2 concentration increases rapidly and the temperature and pressure change at the same time.

In Fig 2.10, observe the curves of the zone having Hazium sensor, we find Hazium curve appear when CO2 concentration accumulating, set points changing, so we conjecture that Hazium may be the product of chemical reaction with CO2 with external conditions, temperature and pressure. Combine the previous patterns, when supply rate rises, air flows fast, the pressure increase, more cold air, lower temperature and vice versa, meets the temperature and pressure. And four sensors we observe, a ventilation environment on floor1, two closed corridors on floor2, a closed room on floor3. The ventilation environment has less change in CO2 concentration and zone temperature, so the Hazium concentration observed is much lower.

 

Fig 2.10    We infer that the Hazium may be the product of chemical reaction of CO2 and it is the harmful chemical produced by the HVAC system.

 

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.

 

Here present the dangerous or serious issues by priority：

1)     Pyoung, facilities member, lost his card around 13:09 on June 1, and reissured his card the next day. It’s strange that there were abnormal records reported by the lost card.

The record was abnormal for arriving other office and recording discontinuous location. Additionally, only information about going to ServerRoom at 10:20 was recorded. For these reason, we speculate that the lost card was picked up and used continually for half a month by ibennett.

 

Fig 3.1      We find the abnormal of pyoung’s card: (A) pyoung001 has quite different radar views so we look into his TimePlaceChart.

(B) His following trace record just between two sensitive places, we infer that ibennett picked his card and take it to the office. In the following days, he sometimes used this card to the ServerRoom.

(C) With map we can find when and where he lose his card. His trace didn’t move from the time in C,. It notes that he lose his card. (D) ibennett came to the place and picked the card back to his office.

(E) pyoung reissued a card, so the former card became black. (F) There were two cards with prefix “pyoung”, pyoung001 and pyoung002, and had independent movement of each other. It was so abnormal.

 

2)     It’s rather curious that Some employees stay at other’s office frequently.

Here shows two examples for verifying the above statement. First, Wreynoso’s location only was detected at btempestad’s office. Second, Acampo often went to Ibarranco’s office. And Acampo even went straight to other’s office without first entering his office on some days. It’s noteworthy and strange that Acampo is able to entering Ibarranco’s office directly in spite of Ibarranco’s absence on June 10th.

 

Fig 3.2                 The phenomena of coming to others’ office quite often

 

3)     There exist missing records at May 31 night.

Employees’ traces ended around zone 1 on the floor 2 at May 31 night so the zone record is discontinuous. We can suppose some unknown reasons caused the prox-card not to work, which can be illustrated in Fig 3.3

 

Fig 3.3      The missing records at May 31st night

 

4)     There are some anomalies about card status, for example, the card was lost, the card was picked up by others.

It’s beneficial to track the card status in the Prox-Map. We find that Gflorez, administration member, lost his card on May 31st, June 1st and Jun 2nd sequentially and he lost his card again on June 7th but got it back quickly because someone pick his card up and put it on the reception office. In general, we can find the card remaining records. For example, Jsanjorge, the CEO, lost his card around 14:00 on June 1st, and got it back later. It seems there was barely any influence on him so that he needn’t worry about how soon he can get his card back. We can infer that some respectable employees are free to access some zones without prox-card so losing their card may become sensitive and even dangerous.

 

Fig 3.4      Track the status of the card, we can see the lost of the card or picked up by others.

 

5)     We find some employees’ actions, including absence, egress, overtime and some anomalies with the RadarView(Fig 1.3).

We find out those specific employees and dates with obvious distinction compared between a day and a week in radar view. Generally, the value of f1z1 axis will become exaggeratedly large when the employee go out or others picked up his lost card. And the absence actually produces a empty radar view. Furthermore, we should observe those persons and the corresponding time so as to determine whether they had asked for a leave or gone out to work.

a)      Mbramar went out at 8:00 on June 9th ,and work overtime on June 11th .

b)     Lcarrara worked overtime on June 5th .

c)      Llagos worked overtime on June 12th .

d)     Rpantanal went out at 12:30 on June 3rd 12:30

e)      Ibarranco went out at 13:01 on June 3rd and was absent on June 10th.

f)        Emintz went out at 14:30

g)      Jsanjorge, the CEO, had records of going out , being absent, working overtime

6)     We find the air-conditioning of the zone 8 on the floor 2 was out of order at June 2nd 6:30.

In Fig 3.5, the CO2 concentration and thermostat temp rose abnormal. The damper position should have turned wider and the flow rate zone supply should have risen, however, they are none. Something was wrong until 15:00 when the damper position and flow rate of  returned normal.

 

Fig 3.5            An abnormal of air conditioning in f2z8 at June 2nd 7:55 and recovered at 15:05

 

7)     We find the air conditioner in z8f3 encounter wrong settings of cool point and heat point during June 9th 10:00 to 22:45.

Fig3.6 presents that the cooling set point and heating set point changed dramatically, which resulted in producing a large mount of reheat power at 10:00 on June 9th,  so that the temperature in the zone was abnormally hot(32℃), and was recovered at 22:45

 

Fig 3.6            Wrong setting of the Cooling and Heating Setpoints caused abnormal temperature

 

8)     Maybe someone has noted that a zone of floor 3 has abnormal temperature through figures above.

The thermostat temperature in the office of the CEO, Jsanjorge, became abnormal from June 2nd 5:00. The Cooling Setpoint and the Heating Setpoint changed dramatically, and the thermostat temperature were not comfortable because of the fact that 20℃ in the morning and 30℃ in the afternoon

 

Fig 3.7      the abnormal temperature and settings of jsanjorge’s office

 

9)     We discover the HVAC system’s Cooling Setpoint and Heating Setpoint become abnormal before June 9th, which started from June 7th on the floor1 and floor2, and June 3rd on the floor3.

It's a fact that the temperature was over 26℃ from 7:00 to 22:00 and below 20℃ at other time.  Besides, the Conf rooms of every floor had a swing at June 1st on Setpoints. According to previous patterns, the consequence of Setpoints fluctuating will be the abnormal reheat power, the abnormal supply air, the abnormal temperature, and the worse quality of the air, CO2 concentration rises, Hazium appears.

Fig 3.8      the wrong change of the Cooling Setpoint and Heating Setpoint caused the abnormal CO2 concentration and temperature

 

10)                     In the first question, we recognize typical patterns for departments, some of which definitively share consistent behavior, so we can detect some anomalous behaviors by comparing the distinction between the polylines in the TimePlaceChart.

For example, administration member lcarrara, almost every day went outside for half an hour at 9:30.

Additionally, it’s an interesting discovery that there are 125 employees registering in employees list while only 113 peoples have prox card records, so we can surmise these employees with no record were on business during the period. And we would be pleasure to see that Security member named “Summer Holiday”.

 

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)     We find the strategy of the HVAC system maintaining the thermostat temperature at one zone. If someone is in the zone, the reheat power would be raised to keep the temperature or else the switch of the air conditioner is preferred to function.

 

The left of Fig 4.1 presents that the HVAC system would retain the indoor constant temperature through the way of switching on-and-off when there was nobody around. When air conditioner came to a halt, the damper and flow rate would disappear, which resulted in declining air quality, such as the augmentation of CO2 concentration. This may cause the problems of the air quality and make people feel uncomfortable. Such as , the abnormal high level of the CO2 concentration. However,the advantage of switching on and off  is energy conserve. So the HVAC system keep the room’s temperature by switching on and off only when there is no one.

 

The right of Fig 4.1 illustrates that there were 6 persons in the zone, so the HVAC system kept the comfortable temperature and suitable flow rate for them through the reheated power heating the supplied cool air.

 

Fig 4.1      The air conditioning maintains the thermostat temperature depending on whether there exists people.

 

 

2)     We find the number of employees in area influence the CO2 concentration and the temperature of that zone. Generally, the more people are in a zone, the higher CO2 concentration and temperature in the zone, and vice versa. The HVAC system would decide work mode according to the number of employees in an area.

 

Looking at the example of Fig 4.2,

(A) show that there were 45 people attending a meeting in the Mtg room(f2z14) at 11:10 on May 31st, so the room hold very high CO2 concentration.

(C) According to the previous findings. The HVAC system was able to enlarge the damper and accelerate the flow rate in the zone, so that the temperature of the zone barely was changed lot.

(D) The CO2 concentration declined until 12:00, when the meeting was over and they went down stairs for lunch.

It’s worth mentioning that the next peak of CO2 concentration was still during next meeting, which is verified by the fact that the CO2 concentration reached a peak again after the  employees returned and convened another meeting at 14:00.

 

Fig 4.2      From the prox and building data, we find relationship between the number of people and CO2 concentration

 

3)     We infer that pyoung001(or lbennett001) entered the ServerRoom and set the inappropriate Setpoints for the building.

 

According to the previous conclusion, the inappropriate Settings of the Cooling Setpoint and Heating Setpoint resulted in the abnormal of HVAC system. It may cause the unreasonable temperature and air problem. Also, we learned that the Setpoints are the same in all other places , so we speculate that some places were capable of adjusting the Setpoints of all places.We suppose the specific place is the ServerRoom.

When we examined carefully the traces passing the ServerRoom, we find that the staff pyoung001 never came to the ServerRoom with other people, and always appeared before the abnormal Setpoints events.Otherwise, we also know lbennett001 picked up the card pyoung001so he should be considered as an important target according our previous conclusion

 

4)     We find the employees of the facilities department can provide help and support when air conditioners need to be adjusted.

 

Looking at the example of Fig 4.3, (A) showed the situation in morning shift. There was no one in this zone at first, so the HVAC system kept the temperature by switching on and off. Acoordingly, a lot of CO2 had been accumulated in the zone. Until earpa and vawelon, the member of the facilities, coming in the zone, the air conditioner modulated its work mechanism to make the air fresher.

 

(B) According to the previous conclusion, we know there was something wrong with the air conditioner in the zone 8 on the floor 3, and it was recovered at 22:45 on June 6th. Because there were total 14 people in the building before the air conditioner resumed normal work, we infer that there exist relationships between air conditioner resuming normal and these people. Furthermore, we find people in f2z1 were all employees of facilities department by inspecting their all identities. For the reason, we can speculate that employees of facilities take on adjusting the air conditioners in the f2z1.

 

(C) In the light of the previous conclusion, there was something wrong with the air conditioner in zone 8 on the floor 2, and it was recovered at 15:00. Conjecturing the possibility that the employees of the facilities department taking on maintaining the air conditioner in f2z1, we deliberately observe the facilities in the TimePlaceChart. As expected we find many traces of f2z1(the conf) around 15:00 definitively

 

Fig 4.3      We find relationship between air conditioners and department of Facilities

 

 

(5) We find there exists some energy wasting.

 

In Fig 4.4(A), Tracking the records of f1z5, we find the lgMtg had always no one except the period from 11:25 to 12:25 on May 31st. However, the air conditioner still worked day and night. This would result in energy wasting. Generally, the number of people in one area should match the power of lights and equipment, and employees should remember to turn lights and equipment off when leaving the office.

 

In Fig 4.4(B), the Engineering and Info Tech had a meeting at 2700 at 14:15 on May 31st. However, there was only one person in the zone but the lights were on in many rooms. We can speculate that a reason of energy wasting is due to forgetting turning off the lights and equipment

Fig 4.4      We find some phenomena of wasting of energy