VAST 2017 MC2 Submission:

Team : Knowledge Based Systems Inc.

Members:

Karthic Madanagopal (primary) (kmadanagopal@kbsi.com)

Kalyanakrishnan Vadakkeveedu(kvadakkeveedu@kbsi.com)

Reuben Fernandes (rpfernandes@kbsi.com)

Akshans Verma (akverma@kbsi.com)

Sanjeev Nookala (snookala@kbsi.com).

Tools used: Tablaue, Matlab

Number of Hours: 40

MC 2.1

Figure 1: Sensor Readings v. Time for All Four Chemicals

Figure 1 shows time series plots of all nine sensors for each of the four chemicals. All sensors seems to be recording readings that vary with time for the three months for which data is available with only a small percentage of missing values. This indicates that the sensors are powered and recording readings at all times. However it is difficult to answer with certainty whether the sensors are recording the correct values from the MC2 dataset. We observe the following interesting patterns: SENSOR 4, seems to be recording values with an increasing trend from April to August to December. Some additional observations about the sensor readings are presented in Figure 2.

Figure 2: Monthly Sensor Reading Trend for Each Chemicals

From Figure 2, we see sensors 4, sensor 5 and sensor 9 show an increasing trend in each of the four the chemical readings. Sensor 1, sensor 2, sensor 7 and sensor 8 show a steady level pattern. Sensor 3 seems to be at an elevated level over the entire duration of data collection. Sensor 6 show a distinct pattern, possibly due to the fact that Sensor 6 is the only sensor situated in the midst of all four factories. Sensor 6 will give some clues regarding the origin of these chemicals for MC 2.3.

MC 2.2

The sensor readings when summed up shows higher levels of the four chemicals in the air with time. This is illustrated in Figure 3. This points to the possibility that the increase in the levels of the four chemicals could be a significant factor in the decrease in the Rose-Crested Blue Pipit population.

Figure 3: Sum of Sensor Readings v. Time for each Chemical

Figure 4-7 show the readings from each of the nine sensors for each of the four chemicals (Figure 4 – AGOC-3A, Figure 5- Appluimonia, Figure 6 – Chlorodinine, Figure 7 – Methylosmolene.) These figures show the sensor reading as a function of the wind destination direction. Wind destination direction is 180 opposite to the wind source direction that was provided in the MC2 dataset. This convention was used to visualize the direction in which chemicals will be carried by wind. Lighter colored sectors in the sensor disc show lower sensor readings and darker colored sectors indicate higher sensor reading values. The four green circles represent the factory locations.

Each sensor reading is depicted as a disk that is broken down into smaller cells by distance from center and azimuth angle. It can be observed that the sensors closer to the factories show higher concentrations of the chemicals (darker sectors present in sensors 3-6, sensor 7 and sensor 9).

Figure 4: AGOC3-A Readings with Wind Directions

Figure 5 shows sensor reading as a function of wind speed and direction away from the factories. Darker cells represent higher readings for Appluimonia.

Figure 5: Appluimonia Readings with Wind Directions

Figure 6 shows sensor reading as a function of wind speed and direction away from the factories. Darker cells represent higher readings for Chlorodinine.

Figure 6: Chlorodinine Readings with Wind Directions

Figure 7 shows the sensors reading as a function of wind speed and direction away from the factories. Darker cells represent higher readings for Methylosmolene.

Figure 7: Methylosmolene Readings with Wind Directions

MC 2.3

Figure 8 shows a set of charts for each of the chemicals as sensor reading versus wind direction and speed. The size of the dot (disc) in the plot represent the magnitude of the sensor reading value. We can observe that AGOC-3A, Chlorodinine and Methylosmolene have many large readings in multiple sensors (e.g., sensors 4, 5 and 6). Sensor 3 seems to be showing readings from all directions. Sensor 3 is not very helpful in resolving the chemical source to one of the four factories.

Figure 8: Sensor Reading v. Wind Direction, Speed & Chemical Type

The wind direction and its effect on the nine sensor readings were used to decipher the factories that are responsible for the observed chemical release. Figures 9-12 show the peak sensor readings as a function of wind direction away from the factories. Each individual cell in the sensor disc shows the sensor value (as dark color), wind speed (as distance from the center) and wind direction away from the source (as azimuth angle of the cell/sector). By examining all the sensors and their direction of its dark patch of colored cells, we were able to estimate the source responsible for the release of each of the chemicals.

Figure 9 shows that for the chemical AGOC-3A, sensor 5, sensor 6 and sensor 9 have dark patches that point away from the direction of the Kasios Office Furniture factory. Sensor 4 dark patch, and the two factories Roadrunner Fitness Electronics and Kasios are located nearly along a line. Therefore, sensor 4 points out that one of the two factories are responsible for the release of AGOC-3A. Therefore, sensor 4 does not resolve between the two factories, however it does not contradict the conclusions from the other four sensors mentioned above).

Figure 9: Peak AGOC-3A Readings as a Function of Wind Destination Direction

Figure 10 shows that for the chemical Appluimonia, sensor 5 and sensor 9 have dark patches that point away from the direction of the Indigo Sol Boards factory. The patches from the other sensors are not significant enough in this plot to identify any other sources (a chemical can have more than one source).

Figure 10: Peak Appluimonia Readings as a Function of Wind Destination Direction

Figure 11 shows that for the chemical Chlorodinine, sensor 4, sensor 5 and sensor 6 have dark patches that point away from the direction of the Roadrunner Fitness Electronics factory. The patches from the other sensors are not significant enough in this plot to identify any other sources (a chemical can have more than one source).

Figure 11: Peak Chlorodinine Readings as a Function of Wind Destination Direction

Figure 12 shows that for the chemical Methylosmolene, sensor 4 and sensor 5 have dark patches that point away from the direction of the Roadrunner Fitness Electronics factory. In the same figure, sensor 6 and sensor 9 have dark patches that point away from the direction of the Kasios Office Furniture factory. The patches from the other sensors are not significant enough in this plot to identify any other sources (a chemical can have more than one source).

Figure 12: Peak Methylosmolene Readings as a Function of Wind Destination Direction