Entry Name: "KBSI-Madanagopal-MC2"

VAST Challenge 2016
Mini-Challenge 2

Team Members:

Karthic Madanagopal, Knowledge Based Systems Inc., kmadanagopal@kbsi.com PRIMARY

Paul Mario Koola, Knowledge Based Systems Inc., pkoola@kbsi.com

John Freeze, Knowledge Based Systems Inc., jfreeze@kbsi.com

Kalyan Vadakkeveedu, Knowledge Based Systems Inc., kvadakkeveedu@kbsi.com

Student Team:

NO

Tools Used:

Custom tools developed by the team in java and Node.js,

MATLAB,

TIBCO Spotfire Desktop,

Tableau,

Microsoft Excel.

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

150 hours

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

YES

Video

https://www.youtube.com/watch?v=4fbHX-XQIZ4&feature=youtu.be

https://www.dropbox.com/sh/ljktbi7sp8p9smj/AACr0OhqgSdrwh9tNLZr4EGIa?dl=0&preview=VAST2016Demonstration.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.

Please watch the video for more details about our visual analytics methods.

Movement pattern 1:

This figure shows the activity pattern of each employee color coded by department. Each tile represents an employee with time of the day as the x-axis and location (floor and zone) as the y-axis. The plots were overlaid for each day onto the same 24 hour axis. Overall, daily repeating patterns of movements were observed for most of the employees; Engineering employees exhibited similar patterns between different individuals. However, Administration employees had much more variation in the movement pattern among individuals (e.g., clais001 and jfrost001)

Movement pattern 2:

The Gantt Chart given above shows the typical activity pattern of a GASTech employee for a given day. The employees’ department is shown by their background color. The slots are colored based on the room he/she is expected to be present. The MC2 dataset does not have room-level proximity information (only floor and zone level information is available). Room-level granularity was estimated from proximity mapping data, employee office assignments, and information about common places.

Movement pattern 3:

This image displays a co-occurrence histogram for all employee pairs as a heat-map. Self-pairings were removed. The x-axis and y-axis shows the employee grouped by department, where each colored pixel represents the co-occurrence frequency between the corresponding employees. The color ranges from 32 rendezvous (red) to 1 rendezvous (blue). The red square pattern about the diagonal on the top left indicates higher frequency of co-occurrences among Engineering employees. Red colored pixels (higher frequency) occur more frequently among employee pairs of the same department.

Movement pattern 4:

This figure shows the time (horizontal axis) and location (vertical axis) coordinates of employees (each circle), separated by day (each tile), according to their departments (color).. The third floor seems to have mixed set of meetings across departments. The second floor has several meeting rooms, each of which appears to be reserved for department-specific meetings. There are relatively much fewer meetings on days of the weekend (6/4, 6/5, 6/11, and 6/12).

Movement pattern 5:

This image shows conflicts between the mobile robot and check-in point data. Repeated entries in the mobile robot data were removed. Around 200 conflicts were observed. First subplot shows the frequency of conflicts by employee name. Nicoloi Cello had the maximum number of conflicts (60).

The second subplot shows a detailed view of the conflicts. The x-axis shows time, and the y-axis shows location. Each color represents an employee and each shape represents proximity scan type (square for mobile and circle for proximity card). The highlighted points show a group of three employees, Marin Onda, Walton Reynoso, and Yuko Finney, who frequently leave their proximity cards in their offices. On all work days Nicoloi Cello is checked in at Floor3, Zone3 but scanned by robot at F3Z4. The same discrepancy was observed for Raphale Faraldo. Since the zones in these cases are adjacent, it is likely that the mobile robot assigns its location to an observed employee in a nearby zone, due to scanning range overlapping multiple proximity zones.

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.

Pattern 1 & 2:

· This plot shows several variables given for the full building. The outside drybulb temperature has similar fluctuations for all days. ElectricDemandPower, HVACDemandPower, and HEATER variables, and the DELIFANPower variables exhibit different behavior on weekends and on weekdays.

· The TotalElectricDemandPower variable shows spikes on the mornings of June 7^th and 8^th (Tuesday and Wednesday), followed by a lower-than-normal value during the day. These spikes correlate with the HVACElectricDemandPower variable but not with the water heater signals.

Pattern 3:

This plot demonstrates the light power for each energy zone in the building. The lights are off in most zones on the weekends, but are never turned off for F_1_Z_3 (the main entrance), or several other common zones (The hallways on each floor).

Pattern 4:

This figure shows the EquipmentPower and SUPPLYINLETMassFlowRate variables for each energy zone. F_3_Z_9 (the server room) draws much more power and has much higher mass flow rate, which matches the higher cooling requirements for this room.

Pattern 5:

This image demonstrates the ThermostatTemp, ThermostatHeatingSetpoint, and ThermostatCoolingSetpoint for each energy zone. For most times the thermostat temperature remained within the bounds defined by the heating and cooling setpoints, except for a few instances where the setpoints were drastically changed. The heating and cooling setpoints are set lower than normal on the mornings of June 7^th and 8^th, coinciding with the power spikes observed in Pattern 2. After morning has passed, the setpoints are set higher than normal, and the temperature increases on these two days.

Pattern 6:

This plot shows the relative Carbon Dioxide concentration at the return outlet for each energy zone. There is a marked increase in the CO₂ concentration during the daytime of June 7^th and 8^th. The lower-power-usage by the HVAC coincides with the increased concentration of CO₂.

Pattern 7:

This image shows the SUPPLYINLETTemperature, VAVREHEATDamperPosition, and REHEATCoilPower for each energy zone. These values are strongly correlated. Additionally, all exhibit increases during the power spikes on June 7^th and 8^th.

Pattern 8:

This image shows the VAV_SYSSUPPLYFAN_FanPower, BATH_EXAUST_FanPower, VAV_SYSHEATINGCOILPower, and VAV_SYSCOOLINGCOILPower variables for each floor. It appears that the VAV_SYSHEATINGCOILPower sensor is dead, as it never reads any value, and is impossibly consistent. The VAV_SYSCOOLINGCOILPower variable spikes in the mornings of June 7^th and 8^th, when the HVAC power also spikes. The VAV_SYSSUPPLYFAN_FanPower variable also exhibits odd behavior on June 7^th and 8^th. The exhaust fan power is similar to the deli fan power (shown in Pattern 1). The bathroom fans are left on throughout the night of weekdays, but is constantly switched on and off during weekends.

Pattern 9:

This figure shows the VAV_SYSSUPPLYFAN_FanPower, VAV_SYSSUPPLYFANOUTLETTemperature, and VAV_SYSSUPPLYFANOUTLETMassFlowRate. The supply fan power and supply fan mass flow rate are highly correlated. This correlation is expected, since the mass flow rate is directly related to the speed of the fan, which is related to the input power. Floor 3 exhibits activity for all three variables on June 7^th and 8^th. This behavior is not seen in floors 1 and 2.

Pattern 10:

This image shows the VAV_SYSAIRLOOPINTLETTemperature, VAV_SYSSUPPLYFANOUTLETTemperature, VAV_SYSAIRLOOPINLETMassFlowRate, and VAV_SYSSUPPLYFANOUTLETMassFlowRate variables for each floor. The air loop inlet and supply fan outlet mass flow rates are similar, but not identical. The temperatures drop and mass flow rates increase during the mornings of June 7^th and 8^th, the time of the HVAC power spikes.

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.

Anomaly 1 & 2:

· TotalElectricDemandPower and HVACElectricDemandPower show spikes on Jun 7^th and 8^th mornings. HVACElectricDemandPower, however goes to a lower than normal weekday value after the spike. The repeating patterns seen on weekdays May 31 – Jun 03 and Jun 13^th, for these two variables (total electric demand power and HVAC electric demand power) are not observed for Jun 7^th and 8^th and weekends. The spikes in TotalElectricDemandPower correlate with the HVACElectricDemandPower variable but not with the water heater signals.

· WaterHeaterGasRate, WaterHeaterTankTemperature, SupplySideIntletTemperature and DELIFANPower show two differing patterns for the weekdays and the weekends. DELIFANPower also shows a slightly different pattern for the two Saturdays Jun 04 and Jun 11, when it is not operational in the day time.

Anomaly 3:

This figure shows the TheromostatTemperature for all energy zones for all days. ThermostatTemperature values are higher than normal for June 7-8. This anomaly coincides with other anomalies observed in (a) HVACElectricDemandPower (morning spikes followed by lower values for the rest of the day) and (b) anomalous cooling and heating setpoints, for the same days (shown in Pattern 5 MC2.2). Anomalous cooling and heating setpoints possibly caused the HVAC system not to operate as normal, keeping the temperature within the usual operating range of 22-27 degrees on weekdays, for most zones.

Anomaly 4:

This figure shows RETURNOUTLETCO2Concentration for all days and energy zones, in two different chart formats. Both plots show higher CO2 concentration on Jun 7^th and 8^th, possibly due to anomalous (non-functioning) HVAC behavior on those days. According to ASHRAE and OSHA standards, 1000ppm is an excessive CO₂ concentration (http://www.engineeringtoolbox.com/co2-comfort-level-d_1024.html). As seen here, the peak reaches above 4000ppm in some zones, for which www.engineeringtoolbox.com suggests “adverse health effects may be expected.” Even normal days exceed the ASHRAE and OSHA standards, potentially causing general drowsiness for employees within certain zones.

Anomaly 5:

· This figure shows heatmaps of REHEATCoilPower, SUPPLYINLETTemperature, VAVREHEATDamperPosition variations for all energy zones across all days. All three signals show a horizontal streak for zone F_3_Z_1 that is different from other zones. This zone includes room 3000, office of the CEO Sten Sanjorge Jr. The fourth plot show F_3_Z_1 ThermostatHeatingSetPoint, ThermostatCoolingSetPoint and ThermostatTemperature. The pattern seems normal for the first two days and then the setpoints show anomalous values for the rest of the days. These setpoints are anomalous because the setpoints are set either too low or too high and do not change during weekends. The actual temperature in the room does not fall within the setpoints most of the time.

· These variables also exhibit anomalies for Jun 7-8 and Jun 11-13. Jun 7-8 anomalies coincide with the anomalies in HVACElectricDemandPower (spikes followed by lower values), possibly because the HVAC system was not operating normally on these days.

· Energy zones in the 3^rd floor show anomalous behavior for all the three signals on Jun 4-6 (bottom square patch).

· SUPPLYINLETTemperature also shows additional anomalies for energy zones (F_1_Z_2, F_1_Z_5) for Jun 7-8, indicating additional HVAC issues for these zones.

Anomaly 6&7:

· This figure shows VAV_SUPPLYFAN_FanPower, BATH_EXHAUST_FanPower, VAV_SYSHEATINGCOILPower and VAV_SYSCOOLINGCOILPower variations for each floor for all days.

· VAV_SYSHEATINGCOILPOWER sensor shows a constant value, possibly due to faulty sensor or data acquisition system.

· BATH_EXHAUST_FanPower shows two differing patterns for the weekdays and weekends; this variable also exhibit repeating on and off behavior at night time.

· VAV_SYSHEATINGCOILPower and VAV_SYSCOOLINGCOILPower signals show different weekday and weekend patterns; they also show anomalous spikes and on-off behavior for the weekdays Jun 7-8, coinciding with other anomalous patterns in the HVAC system, with some lag; these two signals for floor 3 also show anomalous behavior for Jun 4^th; these signals also show anomalous behavior for Jun 11-13.

Anomaly 8:

This figure shows VAV_SYSSUPPLYFAN_FanPower, VAV_SYSSUPPLYFANOUTLETTemperture, and VAV_SYSSUPPLYFANOUTLETMassFlowRate. These variables exhibit two different patterns: one from Tuesday to Friday and another for Saturday and Sunday. However, on Jun 7^th and 8^th they show deviations from their weekday behavior, coinciding with the HVAC anomalies for the same days.

Anomaly 9:

This figure shows VAV_SYSAIRLOOPINLETTemperature, VAC)SYSSUPPLYFANOUTLETTemperature, VAV_SYSAIRLOOPINLETMassFlowRate and VAV_SYSSUPPLYFANOUTLETMassFlowRate variables. The four variables show regular patterns for weekdays and a different pattern for the weekends. However, they show anomalous behavior on Jun 7^th and 8^th, coinciding with anomalous behavior of the HVAC system observed on the same days.

Anomaly 10:

· The top plot shows identical values for WaterHeaterTankTemperature and SupplySideOutletTemperature for all the days. They show two different patterns for the weekdays and the weekends.

· The bottom plot shows constant values for variables SupplySideInletMassFlowRate, LoopTemperatureSchedule, WaterHeaterSetpoint and PumpPower. Constant values (to four decimal places) for SupplySideInletMassFlowRate and PumpPower could be due to faulty sensors or data acquisition systems.

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.

Relationship 1:

This figure shows the server room check-in data (employee ids and check-in time). Our HVAC building data analysis indicated that the heating and cooling setpoints were changed from their normal values immediately before June 7^th, which caused major temperature changes all around the building (all floors and zones). We have made an assumption that the temperature setpoint controls are in the server room. Employees from only three departments accessed the server room: Information Technology, Administration and Facilities. We assume that the Information Technology department’s use of the server room is normal. The last non-IT person to access the server room before the anomalous behavior was Cornelia Lais (administration) on 6^th June 11:00AM. Later, Facilities employees (Patrick Young, Loretta Bennett) accessed the server room on the 8^th and 9^th of June, when the cooling and heating setpoints were changed back to normal. Surprisingly, on the 7^th of June (major HVAC performance change), no non-IT person accessed the server room. We believe the temperature setpoint change is an administration strategy to test how HVAC performed when cooling the building very low on the previous night and maintaining minimal HVAC operations during the day for cost/energy savings. Cornelia Lais accessing the server room on June 6^th supports this hypothesis.

Relationship 2:

This chart shows the temperature for all energy zones. Floors and zones are on the y-axis on left and the time of the day is the x-axis. The minimum observed temperature is shown in blue, the average temperature in green, and the maximum temperature in red. The general pattern is low temperature at night and average temperatures during the day. Days 159 and 160 (June 7^th and 8^th) have abnormally high temperatures relative to the other days at all zones. Floor 3 zone 1 exhibits high temperatures at all times except for 5 hours in the morning. The server room seems to have low temperatures for nearly all days, as expected.

Relationship 3:

This chart shows the CO₂ concentration at all energy zones, with each tile showing a separate date. Floors and zones are on the y-axis and time is on the x-axis. . The minimum observed temperature is shown in blue, the average temperature in green, and the maximum temperature in red. The CO₂ concentration is unusually high on days 159 and 160 (June 7^th and 8^th). A probable reason for this increase is the decreased HVAC operation due to abnormal heating and cooling setpoints. With less HVAC usage, the air is not circulated as frequently, allowing CO₂ respired from employees to accumulate within the building. Additional analysis eliminated several other possible causes, for the following reasons:

1. More people, who breathe out CO₂ – refuted by no noticeable change in magnitude of proximity data (see below)

2. Exhaust not working – refuted by

3. AC not working (may be people breathe out more when perspire)

4. Hazium concentration affecting the HVAC

5. Lower fraction of outside/reused air.

6. Abnormal HVAC settings

Relatioship 4:

This image shows the number of employees in each energy zone through-out the day. The proximity zones of each employee was mapped to a room number (estimated as described in “Movement pattern 2”). The room numbers were then mapped on to energy zones to provide proximity data for energy zones. We observed expected patterns, such as low occupancy at night and during weekends.

Relationship 4:

This figure shows the daily occupancy time for each employee in his/her assigned room (green) and in the break rooms/deli (red). It is highly likely that on the week of June 6^th the human comfort level dropped due to higher-than-normal temperatures and CO₂ levels. The proximity data were analyzed to investigate if there is any change in the employee activity levels. We observed that around 30% of employees seem to have a small increase in their break room and deli usage.

Relationship 5:

This figure shows the aggregated time spent by all employees on each day, in minutes, either in their assigned room or in one of the break rooms (but not any other rooms). On June 7^th and 8^th, the temperature is 4-5 degrees above the normal room temperature and the CO₂ levels were also higher than normal and recommended levels. On June 7^th we do not observe any noticeable change in the employee activity cycle. However, on June 8^th, employees’ activity patterns shift; the lowest value in this chart occurs on June 8^th. The black horizontal line represents the 10^th percentile for all employees over all days.

Relationship 6:

This figure shows the time spent by employees in a department either in their assigned office or in the break room (but not in any other room). Employees of the Executive group seem to spend much less time in the office on June 8^th relative to normal. As expected Facilities group worked noticeably above their average length of time on the 7^th to solve the temperature issue.