Studies On Illumination Calculation Method Based On Artificial Neural Network

The application of Back-Propagation Neutral Network was studied in the illumination calculation passing the research of physics quantities about illumination, the structure of Back-Propagation Neutral Network was given by emulative software Matlab for this problem. Then, the system of software was written with Visual C++ which is software development tool. The test results show that this system has fast and reasonable advantages, self-adaptation and self-study characteristics, the higher accuracy and makes illumination calculation much simpler.The traditional illumination calculation methods includes: the unit capacity method, the utilization factor method and the gradually order calculation method. These methods can still work out problems for the lesser areas and the illuminating undemanding places, but exist an error about 10% or so. But when the illuminative place area needed is bigger, such as business hall, theater, gymnasium, stairs classroom and large lecture hall, traditional illumination calculation methods have the big calculating amounts and the elaborate steps, and hasn't the self-study characteristics and self accumulative design experience. This method has the very great limit.In recent years, the artificial nerve network technique was applied to obtain plenteous results in many fields. The nerve network is a kind of large-scale information to concurrency processing system. It imitates human brain neuronal characteristics and has an information to distribute saving, parallel processing, self-adaptation and self-study characteristics. More importantly, it can map arbitrary functionally function. In the building lighting design, the nerve network view illumination calculation as "black-box" and need not involve concrete calculation process, but also research the correlation importation and exportation. In the new illumination calculation, after experience training good network, as long as calculation requests were provided to the network, the intelligence illumination calculation can be carried out.