ISSS608 2016 17T1 Group10 Proposal

Where there is life, there is hope. Grover 1969

Proposal

Poster

Application

Report

Project Title / Description

This project is based on the VAST Challenge 2016 – Return to Kronos (Mini challenge 2) which involves analysis of a collection of static data about two weeks of GAStech operations data, including building sensor readings and prox-card movement throughout the building. Deliverables include interesting patterns, anomalies/unusual events that are most likely to represent a danger or a serious issue for building operations, and relationships between the proximity card data and building data elements. [1]

Background

In the roughly twenty years that Tethys-based GAStech has been operating a natural gas production site in the island country of Kronos, it has produced remarkable profits and developed strong relationships with the government of Kronos. However, GAStech has not been as successful in demonstrating environmental stewardship. In Jan 2014, a kidnapping incident happened where several employees of GAStech went missing.

After the successful resolution of the 2014 kidnapping at GAStech’s Abila, Kronos office, GAStech officials determined that Abila offices needed a significant upgrade. At the end of 2015, the growing company moved into a new, state-of-the-art three-story building near their previous location. Even though the employee morale rose somewhat with the excitement of the new building, there are still a few disgruntled employees in the company.

The new office is built to the highest energy efficiency standard, but as with any new building, there are still several heating, ventilation, and air conditioning (HVAC) issues to work out. The building is divided into several HVAC zones. Each zone is instrumented with sensors that report building temperatures, heating and cooling system status values, and concentration levels of various chemicals such as carbon dioxide (abbreviated CO2) and Hazium (a recently discovered and possibly dangerous chemical). CEO Sten Sanjorge Jr. has read about Hazium and requested that these sensors be included. However, they are very new and very expensive, so GAStech can afford only a small number of sensors.

With their move into the new building, GAStech also introduced new security procedures, which staff members are not necessarily adopting consistently. Staff members are now required to wear proximity (prox) cards while in the building.

• Passive prox card readers - Individual building zones were covered with prox card readers. The prox card zones do not generally correspond with the HVAC zones. When a prox card passes into a new zone, it is detected and recorded. Most, but not all, areas are still open to staff members even if they forget their prox cards. People are somewhat careless with their prox cards, but some diligent staff members will go to the security desk and pick up a new prox card if their old one is mislaid.

• Rosie the mobile prox sensor - As part of the deal to entice GAStech to move into this new building, the builders included a free robotic mail delivery system, nicknamed Rosie. Rosie travels the halls periodically, moving between floors in a specially designed chute. Rosie is equipped with a mobile prox sensor, which identifies the prox cards in the areas she travels through.

Objectives

As an expert in visual analytics, our team SENSEXiMi STREET have been hired to help GAStech understand its operations data. Two weeks of building and prox sensor data were provided. The challenge is to use visual analytics to identify typical patterns of and issues of concern, as follows:

1. What are the typical patterns in the prox card data?
2. What does a typical day look like for GAStech employees?
3. Describe up to ten of the most interesting patterns that appear in the building data. Describe what is notable about the pattern and explain its possible significance.
4. Describe up to ten notable anomalies or unusual events you see in the data. Prioritize those issues that are most likely to represent a danger or a serious issue for building operations.
   
5. 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.

Notwithstanding the objectives posed in the VAST Challenge, we will be focusing on building suitable web-based dashboards and visualisations that can aid in the analysis of such data. As such, the poster and report will be emphasizing the methodology and visualisations instead of answering the above questions.

Motivations

Analysis of sensor data, or popularly known as sensor analytics, is having a lot of attention nowadays, along with the buzz around big data [2]. A sensor analytics system can help to detect anomalies, and monitor events in real-time [3]. In terms of building management, the use of sensor analytics can help to improve facility’s efficiency and reduce equipment and energy costs dramatically [4].

Sensor analytics can also be fitted with people analytics, to track movements of people. In the context of an office environment, this may be used to track employee effectiveness [5].

Challenges

How to understand the data?

The HVAC data is technical knowledge - from the data variables, it is hard to envisage what the individual items mean. Online research is done to improve comprehension of the subject matter. The following Youtube video provides a general description of the HVAC - https://www.youtube.com/watch?v=fqvo7bSr6t8.

To examine the relationships between the HVAC variables, Tableau dashboards were created to visualise their readings over time, and vis-a-vis other HVAC variables. For better interactivity, the variables were built into parameters, so that users can select whichever variable to examine by just selecting from the drop-down list. The colours of the line charts were deliberately set as from blue (low) to red (high) so that the highs and lows are distinct. Median lines with quantiles were added to provide the analytics for observing anormalies.

• Zone-specific HVAC readings (1st floor) - https://public.tableau.com/profile/lindateo#!/vizhome/HVACzoneF1/Floor1-HVAC
• Zone-specific HVAC readings (2nd floor) - https://public.tableau.com/profile/lindateo#!/vizhome/HVACzoneF2/Floor2-HVAC
• Zone-specific HVAC readings (3rd floor) - https://public.tableau.com/profile/lindateo#!/vizhome/HVACzoneF3/Floor3-HVAC
• Building / Floor HVAC readngs (in 2 separate worksheets) - https://public.tableau.com/profile/lindateo#!/vizhome/HVACbuildingfloor/HVAC-Bldg

Despite research into the subject matter, it still remains technical knowledge on how the different variables interact with one another. Nevertheless, our objective is not to understand how the HVAC works, but rather to identify anomalies. The readings which are identical in patterns are also placed beside each other as a way to "reduce variables". One way is to compare the same day of the week, eg 1st Jun (Wed) to 8th Jun (Wed) as it is assumed that their patterns should be similar on a normal working day. By visualising the readings charts using Tableau, the following preliminary observations were noted:

1. Water Tank Temperature (and Supply Side Outlet Temperature) appeared to spike higher on 4 Jun (Sat) from 7:15pm to 5 Jun (Sun) 6am, then from 5 Jun (Sun) 6:15pm to 6 Jun (Mon) 6am. Although the difference was only around magnitude of 1, it may be raised for concern if the HVAC system is based on precise configurations.
2. Electric Demand Power (both HVAC and Total) recorded sudden spikes on 7 Jun (Tue) at 7:05am, and on 8 Jun (Wed) at 7:05am.
3. Electric Demand Power (both HVAC and Total) from 10 Jun (Fri) 6pm onwards did not follow the previous week's pattern - higher power was recorded until 13 Jun (Mon) 5:50am where it reverted to normal.
4. Some of the VAV readings for 1st and 2nd floor seemed to be off from 5 Jun (Sun) to 6 Jun (Mon), compared to 12 Jun (Sun) and 13 Jun (Mon). 3rd floor was not affected though.
5. Floor-specific readings saw anomalies on 7 Jun (Tue) and 8 Jun (Wed) consistently through all HVAC readings for all 3 floors. Some of the readings continued on the abnormal pattern on 9 Jun (Thu) while the rest reverted to normal.

How to create the visualisations using D3.JS / R?

The learning curve is very steep, and we are also faced with the following issues:

• What visualisation(s) can best represent the findings
• Are we able to find reusable scripts for the selected visualisation(s)
• How do we prepare the data in the format suitable for the working scripts, and what formats were used in the first place (to note that while the scripts were provided in open sources, the formats of the datasets may not be revealed

With advice from Prof and many hours of hardwork and explorations, we managed to develop a few D3 visualisations. We are still in the midst of learning R, from the online resources. Though it may not be in time for the Poster Day, it is still a skill which is worth learning.

Time constraint

This is a note to future VA students - to start picking up Shiny and/or D3 in the earlier weeks. There are lots of resources online, eg Youtube videos and references, hence if given time, you should not have any problems designing a good visualisation.

Outcome

Despite the challenges, we have managed to pull through in one piece. Please look at out poster and analyses in the other tabs, and feel free to let us know your comments. It has been an enriching journey under the guidance of Prof Kam.