Experimental Study On Bio-briquette And Application Of Neural Network & Fractal Theory

As a section of mechanical study on synthetic utilization of biomass energy, which was sponsored by National Science Fund for Distinguished Young Scholars (50025618), this thesis focused on investigation into bio-briquette technology, mainly including experimental study on briquetting character, combustion feature and pore structure, and application of fractal theory, artificial neural network (ANN) and genetic algorithm (GA).The experimental study comprises three parts:1) Experiment on briquetting character: Different briquetting conditions were considered including different biomass and coal, mixture ratio, and moisture, etc. With self-designed module press machine, orthogonal experiment method was applied to organize briquetting and compression test for optimal briquetting conditions.2) Experiment on combustion features: Two combustion experimental systems were established for investigation into sulfur retention rate and combustion properties. With numerical computation, "sight locating" method was raised to determine kindling point.3) Experiment of mercury porosimetry: With the application of mercury porosimetry method, the inner pore system of bio-briquette was measured and analyzed to study characteristic parameters of pore structure and their relationship, and relationship between pore structure and combustion features.With mercury porosimetry experiment and based on Menger manifold, a porous fractal model was established to simulate pore system with self-similarity. The model adapts differences in fractal, pore size distribution, volume vs. pore size and specific surface vs. pore size. Numerical simulation results agree with the experiment data, showing the model is capable of reflecting pore information full-scaled.As the main compositions of bio-briquette, the biomass, coals and addition agent, have various physico-chemical characteristics. Consequently, the characteristic correlation between bio-briquette and its compositions is high nonlinear. And as a result of the limited available data, the extraction of knowledge from experimental data to develop some empirical formula is a formidable task requiring sophisticated modeling techniques as well as human intuition and experience. The use of neural network can alleviate the problem to great extent and reach better prediction effect. Hybrid approaches of a BP and GA neural networks were studied to promote the existing BP model for predicting blending coal's performance due to the similarity to coal blending problem.