Blast furnace condition judgment neural network expert system
1 Neural network-based furnace condition judgment system Neural network and expert system integration can play the advantages of both, overcome the shortcomings of each other. Expert system mainly acquires knowledge from relevant experts (the knowledge of blast furnace expert system is mostly expressed by production rules), since There is no effective means of knowledge updating, so the new model of blast furnace smelting process cannot be obtained in time. In addition, the expert system can not cope with the often incomplete data, which is a major problem of rule-based blast furnace condition judgment expert system. Neural networks use learning algorithms to extract knowledge from training samples, so sufficient, accurate, omnidirectional (fully motivated) training data is needed. However, when developing a blast furnace condition determination system, it is often difficult to obtain data that meets the requirements. For example, the probability of the suspension of the modern blast furnace is very small, only once in a few months or even a year, and in such a long time interval, many national key science and technology projects (N.97-56202-02) refine the environmental parameters and Unmeasurable parameters have changed, making it almost impossible to collect an effective, sufficient set of suspension learning samples. However, for the ironmaking process of the furnace, the consequences of the abnormal condition of the suspended material are very serious and will have a large adverse impact. Therefore, the system must be able to judge and forecast. Therefore, it is insurmountable to simply use the neural network to predict the blast furnace condition. defect. In order to utilize the learning function of the neural network and the characteristics of incomplete data, a neural network-based integration method is usually used to judge the blast furnace condition. The establishment process of the neural network blast furnace condition prediction and judgment expert system is as follows: Firstly, the furnace condition judgment and reasoning network is established according to the smelting principle and operation practice; then the reasoning network is converted into the connection network; finally, the signs of abnormal furnace conditions appear in multiple sets of blast furnaces. And the accident category as a model pair, the weight matrix is ​​obtained by the neural network learning algorithm. 14. Using the BP learning algorithm, the system can also complete the refinement of the membership function and the fuzzy rules. The expert system based on neural network uses the self-learning function of neural network to make up for the deficiency of expert system knowledge acquisition, but whether the operation effect can meet the requirements of blast furnace production depends on the generalization characteristics and adaptive characteristics of the system. The generalization feature refers to the ability of a well-learned network to function effectively within the scope of the original instance but beyond the original instance. For those unusual conditions (such as suspension) where the probability of occurrence is small, this attribute reflects whether it can continue to correctly judge the level of adaptive characteristics refers to an ability to adapt to the environment. If a system has adaptive capabilities, the system can change its intrinsic function and adapt to environmental changes. The design trains the neural network as shown. The original developed expert system was run in real time, but this study required a large amount of data input into the expert system in batches, for which a random data generator was developed. The function of the random data generator is to generate random data such as air volume and wind temperature according to the mean and variance of the production data. Run the developed rule-based blast furnace condition judgment expert system, calculate the membership degree, and complete the reasoning. The result of the inference is used as the ideal output to form the training sample. Finally, the self-learning function of the neural network is used to obtain the weight and threshold parameters of the neural network. The functions of each part are as follows. Production data preprocessing. There are interference signals such as noise and abnormal mutation in the original data, and the original data needs to be preprocessed. The preprocessing includes smoothing, calculation of statistical information, and the like. Neural networks and self-learning. This paper chooses a 3-layer feedforward neural network and uses BP learning algorithm. Calculate membership. The membership function is read from the dynamic membership function library, and the control parameters or process parameters are read from the production data, and the membership degree is calculated, and the calculation result is stored in the fact database. Abnormal alarm. When the furnace condition judgment system predicts an abnormal furnace condition, an abnormal alarm program is triggered. The alarm program plays the corresponding recording according to the type of abnormal furnace condition. Accident recall. When the system predicts abnormal furnace conditions, or judges that an abnormal furnace condition has occurred according to the production data, the production data and facts (including original facts and inference results) of the first half hour are recorded in the buffer to prepare the tracking system. Used during operation and analysis of furnace conditions. In addition, when the predicted abnormal condition and the actual abnormal condition exceed a certain error, the system will store the relevant information into 2 graduates. The application of artificial intelligence and expert systems in the steel industry. Journal of Wuhan Iron and Steel University, 1995, 18 (2): 146 4 Yang Shangbao, Yang Tianqi, Dong Yicheng. Neural network blast furnace condition prediction and judgment expert system. Journal of University of Science and Technology Beijing, 996, 18, (3): 220 5 Liu Jinxi, Wang Shuqing. Neural network expert system for abnormal furnace conditions of the furnace. Journal of Iron and Steel Research, 1998, 10 (3): 67 8 China Software Industry Association Artificial Intelligence Association. Artificial Intelligence Dictionary\
Leisure chair is suitable for office and home, hospital, shopping mall and school open space and so on. Finally it depends on the design and style for the space.
High back leisure chair with a swivel base . chair and ottoman setup will give your living room a classic feel and look . Hight back leisure is not only perfect in the home, but makes for a great addition in the office .
Leisure chair with a swivel base: it elevated degree of comfort . ideal for relaxation . comfortable design .One series of classic central cross-shaped chrome base allows 360 degree swivelling . modern charm with an aesthetic appeal . slim volumes and flowing surfaces . swivel base with four spokes, combines elegance and comfort . superb in any environment, from living rooms to reception areas . suitable for contract and project use .
Leisure Chair,Lounge Chair,Sunday Leisure Chair,Home Office Furniture Foshan Hollin Furniture Co.,Ltd , https://www.china-hollin.com