大型顶吹平炉冶炼硅钢

大型顶吹平炉投产以来，由于降碳速度快，因而造渣困难，脱硫能力差，长期不能冶炼硅钢。针对上述问题，对造渣工艺进行改革，获得成功，不仅降低了消耗，提高了质量，缩短了熔炼时间，而且也满足了用户对硅钢的要求，同时，取得了较大的经济效益。     

控冷电磁站搬迁方案的改进

通过大量的调查研究，查清了电磁站内渗漏水的缺陷，提出了解决方案和具体操作方法，改变了原定的搬迁方案，不仅彻底地消除了缺陷，还节约了大修费，缩短了工期，为生产创了可观的效益。     

承钢“八五”期间科技进步及效益综述

承钢“八五”期间坚持“科技是第一生产力”的思想，紧紧依靠技术进步，狠抓节能降耗，增产增收，投资７亿余元用于科技进步的项目，实现了高炉喷吹煤粉，连铸，连轧，热电联产等工艺技术并已取得了显著的经济效益，经济技术指标水平明显提高，有力地促进了企业的发展，提高了市场竞争的能力，保证了经济效益的提高，为“九五”的进一步发展奠定了基础。     

10t"O"型龙门吊的改进与探索

本文介绍10t“0”型龙门吊通过静定车架结构代替传统的超静定结构的改进，其造型美观，受力合理，降低了精度要求，节省了大量的材料，降低了成本，使制造工艺更加方便，大大降低了金属结构的自重，改善了起重机的运行状态，延长了驱动车轮的寿命，故障率低，维修费用少。     

煤气加压站系统研究与优化

热电厂煤气加压站经过系统优化改造后，设备运行可靠，操作方便，满足了生产工艺要求，实现了变频调速，消除了煤气的泄漏现象，达到了节电，节油，节约维修费用的目的，经济效益可观。     

铝-钪(2%)合金的制备

传统的生产铝钪合金的方法，成本高，条件复杂，不利于工业生产，我们对金属热还原制取A1—Sc合金进行了研究，简化了生产条件，降低了对原料的要求，缩短了工艺流程，降低了生产铝钪合金的成本，开拓了铝钪合金的应用市场。     

焦化厂回收易地大修暨脱硫脱氰工程简介

本文叙述了国内先进的煤气净化工艺和氨、硫回收装置，净化了焦炉煤气，减少了环境污染，改善了硫铵质量，增加了硫磺新产品，提高了生产管理水平，确保了安全生产。     

铝冶炼供电系统节能降耗的措施

湖北丹江哆嗦业公司三期技改工程供电整流系统采用了较多的新技术，新设备，新工艺，在该前提下，对运行参数进行详细研究，达到合理利用电能的目的，同时采用全面质量管理工作的方法，成功地改善了电流的稳定性，减少了设备停运率，为安全生产提供了有力的保障，达到了节能降耗的目的，取得了一定的经济效益。     

基于问题原型的采矿方法选择专家系统的研究

本文针对矿山的实际需要，以矿山工作人员应用对象，在充分分析矿块回采问题的基础上，提出了三种问题原型，建立了领域问题的知识库，编制了推理机，用户接口，解释，知识库管理等程序模块，形成了采矿专家系统的框架，为进一步开发奠定了基础。     

混合轻稀土萃取分离工艺优化研究

针对现行稀土萃取分离工艺存在的问题，对混合轻稀土萃取分离工艺进行了系统的优化研究，以模糊分组新工艺取代了传统分离工艺，并对模糊分组新工艺与传统分离工艺的技术经济指标进行了全面的比较，结果表明，模糊分组新工艺，分离效果好，大幅度降低了酸，碱消耗，减少了有机相和稀土的存槽量，既降低了生产成本，又节省了充槽费。     

Neural nets – types,configurations and pitfalls

Fuzzy logic, neural nets and genetic algorithms form the core of soft computing methods. They are useful when there is no possibility to compute an exact mathematical model (hard computing). Neural nets have the ability to learn by example. This advantage is exploited by a lot of applications and many software packages make it quite easy to use neural nets. A stage is reached, where some critical remarks should be made in order to avoid disappointments. Some frequently used net types (backpropagation, LVQ, SOM) are discussed together with configuration and training problems. Important topics are the avoidance of under- and overfit and the remark that neural nets produce correct outputs only if the inputs lie in the part of the feature space, the net was trained for. Therefore, a detailed analysis of the training data set should be made. In the context of safety relevant applications the missing interpretability of neural net outputs is often criticized. Fuzzy-neuro-systems try to improve this situation.     

Fast sigmoidal networks via spiking neurons

We show that networks of relatively realistic mathematical models for biological neurons in principle can simulate arbitrary feedforward sigmoidal neural nets in a way that has previously not been considered. This new approach is based on temporal coding by single spikes (respectively by the timing of synchronous firing in pools of neurons) rather than on the traditional interpretation of analog variables in terms of firing rates. The resulting new simulation is substantially faster and hence more consistent with experimental results about the maximal speed of information processing in cortical neural systems. As a consequence we can show that networks of noisy spiking neurons are "universal approximators" in the sense that they can approximate with regard to temporal coding any given continuous function of several variables. This result holds for a fairly large class of schemes for coding analog variables by firing times of spiking neurons. This new proposal for the possible organization of computations in networks of spiking neurons systems has some interesting consequences for the type of learning rules that would be needed to explain the self-organization of such networks. Finally, the fast and noise-robust implementation of sigmoidal neural nets by temporal coding points to possible new ways of implementing feedforward and recurrent sigmoidal neural nets with pulse stream VLSI.     

Survival analysis and neural nets

We consider feed-forward neural nets and their relation to regression models for survival data. We show how the back-propagation algorithm may be used to obtain maximum likelihood estimates in certain standard regression models for survival data, as well as in various generalizations of these. Examples concerning malignant melanoma and post-partum amenorrhoea during lactation are used as illustration. We conclude that although problems with the substantial number of parameters and their interpretation remain, the feed-forward neural network models are flexible extensions to the standard regression models and thereby candidates for use in prediction and exploratory analyses in larger data sets.     

Automatic grain size determination and classification of iron carbides with neural nets

Under the aspect of quality assurance material testing is getting more and more important. As a consequence there is a strong demand for automation and objective procedures. This article concentrates on the inspection of polished steel specimens using image processing and intelligent software techniques. Nowadays computer aided image processing of microstructures is widely used, but final analysis and quality grading are based on the experience of the testing personnel. Using intelligent software techniques (soft computing), especially neural nets, allows complex decision processes to be done on a computer in a reproducible manner. Of course there are certain premises: suitable restriction of the application domain, careful image preprocessing and feature extraction, critical surveillance of the training process. The article focuses on two neural net applications: grain size determination and classification of iron carbides. To learn the ability to do complex decisions in a reproducible manner from a suitable example data set – at least in a restricted domain – is the main advantage of a neural net solution.     

Modal Seismic Control of Building Frames by Artificial Neural Network

An artificial neural network based control scheme for reduction of the seismic response of a multistory building frame is presented. The scheme, which controls the response of the frame by also controlling the significant modal contributions to the overall response, is designed to provide a target reduction of response and to take into account the time-delay effect. Inputs to the control scheme are (1) the measured accelerations at only a few selected points of the structure and (2) the ground acceleration. For developing the control scheme, two sets of neural nets are trained. The effectiveness of the control scheme is tested for both known and unknown problems from the El Centro earthquake record for a 10-story building frame. Results of the study show that the control scheme is highly effective in controlling both displacement and acceleration responses of the frame for the unknown El Centro earthquake record.     

Adaptive neural net (ANN) models for desulphurization of hot metal and steel

Neural nets can be adapted to complex patterns of interrelated input and output variables in a process even if the data sets contain a lot of noise. In this work two specific examples for the application of adaptive neural nets (ANN) in steel Industry are described. First, the sulphur content of hot-metal, obtained at the end of calcium carbide powder injection into 4001 torpedo ladles is predicted as a function of hot-metal weight, treatment time, initial sulphur content, gas flow rate and powder injection rate. The values predicted by the trained ANN model for a completely new set of input test data compare well with the actual values obtained on the shop floor. In the second example, the sulphur content of steel, obtained at the end of blow is predicted as a function of liquid-metal weight, total amount of oxygen injected, amount of iron ore added, and the temperature, contents of carbon, manganese, phosphorus and sulphur determined by in-blow sampling in a 300 t converter. The ANN predicted values of sulphur content of steel at tap (without reblow) also agree well with the values obtained on the shop floor.     

Modeling of metal-slag equilibrium processes using neural nets

Although numerous methods have been proposed in the literature for the modeling of slag/metal equilibrium processes, these models are usually semiempirical hybrids of regression equations and thermodynamic equilibria. Fundamental models are successful only in a few cases, owing to the complexity of interaction between variables in real pyrometallurgical systems. This article shows that neural networks are suitable for modeling such ill-defined equilibrium processes, with noa priori knowledge being required about the form of any model. A three-layered perceptron or backpropagation network is used in this article. A weight matrix describes the relative strength of the interconnections between the network nodes and therefore the relative importance of variables. A number of sets of experimental data could be used to train the neural net which can then be used to predict other sets of data. Neural nets have been used successfully for the modeling of activities in metals and slags, the distribution of species between metal and slag, and slag viscosity on the basis of published data.     

对邢钢北京世贸商城工程冷轧带肋钢筋焊接网设计分析

对邢钢北京世贸商城工程的冷轧带肋钢筋焊接网设计进行了分析,认为该工程各部分的底网设计、纵向面网设计和横向面网设计合理,针对该工程编写的施工方案,在实际施工中取得了良好的效果。     

Analytical Construction of Transient Flow Nets in Homogeneous and Isotropic Flow Medium

The purpose of this study is to construct time-dependent flow nets, also called transient flow nets in homogenous and isotropic flow medium. Transient flow nets under hydraulic structures are developed in response to reservoir head fluctuations. An analytical solution for a transient flow net has not been reported in the literature. Time-dependent flow net equations are limited in engineering applications to simple boundary conditions. The geometry of transient flow nets does not change with time, as only the numerical values assigned to equipotential lines and flow lines change with time.