Use of artificial neural network process analyzers: A case study

In this paper artificial neural network (ANN), which are known for ability to model non linear system, are used as his analyzer to predict his in export gas at South Pars Gas company in south of Iran. The training data has been collected over duration of three years and use to train a neural network to develop based hrs analyzer. The results are promising. The FPD become the most commonly used sulfur detector due to its ease-of-use and relatively low cost of operation. It also exhibits moderate sensitivity and selectivity for sulfur. However, the FPD has a number of disadvantages, a key disadvantages is that its response to sulfur is approximately proportional to the square of the mass of the sulfur compound. Nowadays PFDD used as sulfur detector, it is a significant improvement because it can provide better sensitivity and selectivity for sulfur and phosphorous. But use of PFDD increasing both costs and complexity. The adopted neural network used is a multi layer feed forward, network with 0 Inputs, one output trained using back propagation algorithm. After many trails, it has been found 10 neurons in the hidden layer give an acceptable compromise between network estimation accuracy and network complexity.