City Research Online

Artificial intelligence applications in waste water monitoring for industrial purposes

Benjathapanum, N. (1995). Artificial intelligence applications in waste water monitoring for industrial purposes. (Unpublished Doctoral thesis, City, University of London)


This thesis reports on work carried out in the development of software for artificial intelligence sensing systems based on UV-Vis spectroscopy, designed for remote on-line and real-time analysis for monitoring of industrial effluent. A feasibility study on artificial intelligence methods and the design of an intelligent monitoring system has been researched. This system is capable of detecting the occurrences of chemical pollutants and the concentration of species involved.

The controlling software was developed in this work for the remote modem control of a computer controlled UV-Vis spectrometer system. This provides facilities for signal processing, data storage, and transfer of data to a host machine for real-time analysis. This gives significant advantages in term of automatically and instantly reporting of a pollution incident. This front end sensing system has been installed at industrial sites in order to demonstrate the apparatus in the real situations and to obtain data for qualitative analysis.

Difficulties in working with the above data pointed to the need for a laboratory-based evaluation and modelling of analysis methods. This evaluation and development of suitable methods forms a major part of the work. The samples prepared for a set of data were mixtures of nitrate, hypochlorite and ammonia in various concentrations representatives of that expected in real outflows. This data set presented several significant problems in data analysis, including an overlap of UV absorption bands and the interaction between ammonia and hypochlorite to form monochloramine which has its own specific spectral features.

In the evaluation, the spectral data obtained were analysed by two different methods. The first was Principle Component Analysis (PCA) which is based on linear multivariate analysis, and samples were investigated to compare the effects of interactions between components. The second method was Neural Network analysis, which is a non-linear analysis technique. After considerable effort, this approach resulted in a data analysis scheme where the Back-Propagation algorithm was used as a two-step process. In the first-step, the network inputs were derived by binary encoding segments of the second derivative of the absorption spectra according to their shape and the network outputs specified according to which species were likely to occur. As a result, the second-step network could then focus on a few inputs that strongly correlate with the presence of the expected species. Also the second-step provided a filter that compensated for false classification of species, at low concentration levels. The resulting data analysis scheme depends on a knowledge of the expected chemistry for implementation: however it gives a much better performance than PCA in this particular case.

The complete monitoring system has been integrated with a Graphical User Interface software to perform real time analysis at a host machine. A multi-task system for on-line monitoring of data transmitted from a remote site, has been developed, based on the neural network approach.

Finally, the intelligent monitoring system is demonstrated and evaluated.

Publication Type: Thesis (Doctoral)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments: School of Science & Technology
School of Science & Technology > Computer Science
School of Science & Technology > School of Science & Technology Doctoral Theses
[thumbnail of Benjathapanum thesis 1995 PDF-A.pdf]
Text - Accepted Version
Download (10MB) | Preview


Add to AnyAdd to TwitterAdd to FacebookAdd to LinkedinAdd to PinterestAdd to Email


Downloads per month over past year

View more statistics

Actions (login required)

Admin Login Admin Login