Skip to main content

Section 2: Testing the algorithm using existing data

 pipe

 

Data from available projects is based on different manhole scenarios; varying manhole diameter, relative inlet and outlet pipe angles, and relative inlet and outlet pipe elevations. In order to validate the accuracy and effectiveness of the algorithm, it is planned to test and analyse over 1000 data sets.

 

matlab deconv screenshot

 

By modifying the basic code of the algorithm, the program has been changed to test multiple data sets sequentially. Each iteration of the program outputs a .csv file, as well as a .tiff image file illustrating a graph similar to the graph below. Further modifications will be made to also output a single spreadsheet for an entire group of data sets stating important features of each data set, including the R 2 values.

graph one

The graph shows that the inputs and outputs of collected data for each of the five repeats show good similarity, while the predicted outputs from the algorithm experience noticeable variability.

 

 graph2

This graph illustrates an example of the algorithm failing to accurately predict the output. The main aim for this section of the project is to find the reason for this anomaly. The variance is shown on the purple line.

grpah3

This graph illustrates a successful output from the Matlab algorithm, with comparisons to present modelling techniques, ADE and ADZ predictions, showing a vast improvement on output prediction from the system.

 

grpah4

The time for the system can be normalised so that all the results can be plotted on the same graph and then compared. Using this, a graph showing the cumulative concentration over time can be plotted. The cumulative graph for 3 of the repeated tests is shown.

outflowpic