PRS - back to the homepage   Pattern Recognition Systems - mail us {short description of image}

08.05.04 mail uswebmaster

PRS as
MIX-M?hlenpris PB 24
5006 Bergen
+47 5532 5221
+47 5558 9496

Selected Projects  PRS offer high-qualified consultancy

Automated prediction of a biological parameter using data rich instrumental measurements
This work was done together with Statoil.
Combustion of fossil fuels produces hundreds of compounds. Many of these are mutagenic. The mutagenicity is traditionally measured through bacteriological tests, which are expensive and time consuming. Cheaper and faster chemical analysis can often replace the biological measurements. In this project, full scan GCMS data for a large number of samples were recorded. A method was developed to automatically resolve the GCMS data into the individual components. The resolved data was used to accurately predict the mutagenicity. Identification of the individual compounds in a sample, as well as their contribution to the predicted mutagenicity, is possible.

Identification of critical steps in an industrial process.
This work was done together with Elkem. There is a potential for product quality improvement in many complex industrial processes. However, since there are many steps in a process, it may be difficult to identify the variables with largest impact on the quality. In this project, a strategy was devised to separate the impact of raw material variations from the influence of variation in the manufacturing process.

Fault detection in a dynamic process.
This work was done together with Norsk Hydro. Modern industrial processes are monitored and controlled by advanced and automated process sensors and models. Although this makes it easy to detect that the process is deviating from the optimal operating region, a more thorough diagnosis is often difficult to achieve. There can be errors in the sensor outputs. The real value of a control variable may differ from the apparent set point value. The process itself may have changed, so that the underlying model is no longer valid. Finally, changes in the noise level of the sensors occur. General diagnostic tools for all these situations were developed.

This work was done together with Nutreco. One of our consultants has been working full-time with Nutreco. Performance of processes and quality measurements have been improved using multivariate techniques. Training of staff in multivariate design and analysis has also been provided.
Assessing the reservoir potential in an exploration area.
This work was done together with Aker Geo Petroleum Services.
Identifying the areas with the best reservoir potential is crucial when applying for new exploration licenses. This decision is based on seismic data of the whole area and well-logs and core samples from existing wells in the area. The most important reservoir properties are porosity and permeability, which can only be measured from core samples. In this project, porosity and permeability were predicted from existing well-logs using multivariate models. These predictions were then used to create maps displaying porosity and permeability for the neighbouring unexplored areas.

Environmental survey. Quantitative assessment of the environmental impact from industrial installations.
This work was done together with Norsk Hydro Oil & Gas. The exploration and production of oil and gas in the North Sea disturbs the ecological system on the ocean floor. Biological and chemical data were used to build multivariate models for monitoring the disturbance of the benthic fauna. A multivariate index was created to establish quantitative boundaries between natural variation and variation due to contamination. The results can be utilised to assist the design of future surveys.

Precision farming. On-line models for optimal fertilising of fields.
This work was done together with Norsk Hydro Agri. To maximise the harvest with minimum use of fertiliser is a priority task both from an environmental and an economical point of view. Demand of fertiliser depends mainly on chlorophyll content and biomass. Robust and accurate on-line multivariate models were developed for predicting chlorophyll content and biomass from optical sensors. The models are valid for varying weather conditions. The developed models were able to predict chlorophyll content with an error of 5 percent of the true value.

If you either send us an e-mail or give us a call where you try to describe your current local challenge, we will submit the following information:

*areas where we are able to assist you
*our way of working
*some successful projects

----------------------------Top of Page

© 1997-00 PRS AS. All rights reserved.