BSc (Eng) Civil Engineering Research Project by
TANYARADZWA USHAMBA
DATA COLLECTION
The dam measurements were recorded by relevant trained personnel on site. The different static measurements taken were captured on excel spreadsheets and sent to the University of Cape Town’s Civil Engineering Department.​
Crack width gauge measurements
A total of forty six crack width gauges were installed on each side of the dam wall, making the total number of crack metres on the dam ninety two. The crack metre readings were taken at least once every month from the year 2002 until 2010.
Geodetic survey measurements
The geodetic surveys were carried out twice each year during the winter and summer seasons. In winter, they were mostly taken during the month of August and in summer they were taken mostly in February.
Tilt metre measurements
There are currently eight tilt metres installed on the dam wall. The tilt measurements were recorded since 1986. The measurements were recorded at least once a month throughout each year.
Data cleaning and sorting
Statistical analysis
One of the most important steps during data analysis is data cleaning. In order to get meaningful results from the data analysis, the raw data had to be cleaned. Data cleaning which is also referred to as data cleansing or scrubbing is a process whereby errors and inconsistencies in the recorded data are identified and removed with the objective of improving the quality of the data (Rahm & Hai Do, 2000). This process was done to remove data capturing errors and other invalid data. It was used to recognize and remove outliers in the measurements.
​
In order to analyse the performance of the dam over the years more effectively and easily, the data was sorted into a meaningful order. After cleaning the data collected from the crack metres and replacing missing values, the data was sorted. The measurements were then categorized into seasonal categories namely summer, autumn, winter and spring.
Statistical methods were used to analyse the data. A useful and frequently used method of calculating the degree to which two variables are interdependent is the correlation coefficient. The correlation coefficients of the water level and displacements were calculated for all measurements using Statistical Package for the Social Sciences (SPSS).
The bivariate correlation coefficients were calculated for the crack metre, tilt gauge and geodetic survey measurements. The Spearman correlation coefficient which is non parametric was calculated instead of the most commonly used parametric Pearson Correlation Coefficient because it makes fewer assumptions of the data. A two tailed test of significance of the calculated correlation coefficients was carried out in order to test the significance of the correlations using SPSS.
​
Graph plotting
Graphs of selected targets displacements were plotted against the water level. This was done for the x, y and z displacements.