基于案例推理的钢水温度在线管控模型

钢水温度的精确管控有利于提高铸坯质量和降低生产成本。针对目前炼钢—连铸区段钢水温度在线管控方面存在的不足,在分析钢水温度影响因素的基础上,建立了基于案例推理的炼钢—连铸区段钢水温度在线管控模型。同时通过调整案例推理算法的相似度计算方法、权重计算方法、重用案例个数等参数提高模型的精度。结果表明:转炉出钢温度预定模型平均降低转炉出钢温度6℃,精炼结束温度预定模型提高连铸开浇温度命中率2.33%。精炼开始温度,精炼结束温度和连铸开浇温度预报模型误差小于10℃的命中率分别达到75.33%、98.33%和95.67%,且均高于神经网络模型。     

智能LF控制系统的研究

提出一种基于人工神经网络的智能LF控制系统模型,该系统的核心是LF精炼过程钢水温度和成分的实时预测,并确定了一种部分反馈的人工神经网络进行钢水温度和成分的实时预测。     

RH精炼钢水温度预报模型

根据RH精炼过程钢水传热的物理模型,提出了预报RH精炼过程钢水温度的数学模型.根据此模型可以分别计算真空室内和钢包内钢水的温度变化.计算结果表明:在精炼初期2～3 min内,由于炉壁吸热,钢包内钢水温度迅速降低,随后真空室内钢水温度开始回升,4 min后,钢包和真空室内钢水温度趋于均匀.采用该模型预报实际真空脱碳终了温度,其中平均误差在±5 ℃以内的占72 %.     

基于数据驱动的转炉二吹阶段钢水温度动态预测模型

转炉钢水温度是转炉终点控制的工艺参数之一,精确的钢水温度预测对转炉终点控制具有重要的指导意义。然而,以往的大多数转炉终点预测模型属于静态模型,只能够实现对转炉吹炼终点钢水温度的预测,无法实现动态预测,导致模型的作用有限。针对该问题,提出了一种基于数据驱动的转炉二吹阶段钢水温度动态预测模型。模型先通过新案例主吹阶段的工艺参数,基于案例推理算法找到历史案例库中相似案例。再利用相似案例的二吹阶段工艺参数并基于长短期记忆网络（Long short-term memory,LSTM）算法训练工艺参数与钢水温度的变化关系。然后利用训练好的LSTM模型,计算新案例二吹阶段的钢水温度变化。最后,利用某钢厂实际生产数据,研究了不同重用案例个数及神经元个数对模型预测精度的影响,实验结果表明:模型在重用案例个数为4,神经元个数为10时模型的预测精度最高,此时模型对钢水温度的预测误差在[?5 ℃, 5 ℃]、[?10 ℃,10 ℃]和[?15 ℃,15 ℃]的命中率分别达到40.33%、68.92%和88.33%,模型的性能高于传统二次方模型和三次方模型。      

LF炉精炼过程中钢水温度变化研究

对LF精炼过程中影响温度变化的主要因素(包括LF精炼进站温度、通电时间、软吹时间、精炼总处理时间、石灰加入量及Al加入量)进行了统计分析,并在此基础上利用多元线性回归方法建立了钢水温度变化模型。将实际生产过程中相关数据采用模型进行计算,计算温度与实际温度误差在±10℃的比例达到了86%以上,对现场生产具有一定的指导作用。通过对模型进行分析得知,引起精炼过程中温度降低的因素依次为精炼时间、炉料加入量及软吹时间。且通过优化吹氩制度、缩短精炼时间可以有效的减少钢水温度损失。     

LF钢包精炼过程控制模型开发

根据冶金机理和钢液热平衡原理,综合考虑电极供电、合金、渣料添加情况及钢包状况等因素,并结合人工神经网络算法和自学习方法,建立了LF精炼过程控制模型,包括温度预报模型和合金模型。经测试,温度预报模型在线计算值与实测值的绝对偏差在±5℃以内的比例达到了85%以上,合金模型应用后,LF钢水成分合格率从90.1%提升至94.7%,平均合金化时间由8.9 min降低到6.7 min。利用Microsoft Visual.Basic.net程序设计软件以及Microsoft SQL Server 2000数据库软件,开发了LF精炼过程控制模型软件。     

RH-MFB二次精炼过程钢水温度控制的模拟

RH-MFB二次精炼过程钢包钢水传热行为和钢水温度的变化规律,建立钢水温度预报模型,编制了计算机软件对实际过程进行模拟。通过对9炉300 t钢水RH-MFB精炼时钢水温度的预测结果表 明,钢水温度的计算值和实测值的误差小于±10℃。按目标温度要求,进行补偿措施,有效地控制钢水温度     

RH-MFB精炼时钢水温度预报模型的开发

分析研究了RH-MFB精炼时脱碳过程、脱氧、合金化、吹氧加铝、非操作因素对钢水温度的影响,并建立了精炼钢水温度预报模型。通过对连续精炼的10炉270 t超低碳钢水(0.001%~0.0025%C)温度的验证结果表明,模型计算温度和实测温度的误差不大于±5℃。     

基于BP神经网络的钢水温度预定模型

针对目前钢水温度预定方法存在不足,在分析钢水温度预定原理的基础上,在邯钢邯宝炼钢厂建立了基于BP神经网络的精炼终点目标温度和转炉终点目标温度的动态预定模型。利用邯宝炼钢厂的历史生产数据对模型进行了训练和测试,并进行了现场应用试验。结果表明,预定模型对转炉和精炼终点目标温度进行了优化,应用预定模型后,LF开始温度命中率提高到75%,中间包温度命中率提高到96.7%。     

马钢CSP流程RH精炼钢水温度预测模型

基于冶金机理和传热学计算,分析研究了RH精炼过程中脱碳、吹氧加铝、脱氧、合金化、喷粉、真空室状态以及钢包等级等各类因素对钢水温度的影响。结合现场实际生产数据,建立了RH精炼钢水温度预测模型,经过对实际生产跟踪验证表明,模型预测的钢水终点温度与实测值偏差在±5 ℃以内的命中率为87.42%,偏差在±8 ℃以内的命中率为100%。     

Comparison of Case-Based Reasoning and Artificial Neural Networks

The outcome of construction litigation depends on a large number of factors. To predict the outcome of such litigation is difficult because of the complex interrelationships between these many factors. Two attempts are reported in the literature that use, respectively, case-based reasoning (CBR) and artificial neural networks (ANN) to overcome this difficulty. These studies were conducted by using the same 102 Illinois circuit court cases; an additional 12 cases were used for testing. Prediction rates of 83% in the CBR study and 67% in the ANN study were obtained. In this paper, CBR and ANN are compared, and their advantages and disadvantages are discussed in light of these two studies. It appears that CBR is more flexible when the system is updated with new cases, has better explanation facilities, and handles missing data and a large number of features better than ANN in this domain. If the use of CBR and ANN is understood better and if, as a result, the outcome of construction litigation can be predicted with reasonable accuracy and reliability, all parties involved in the construction process could save considerable money and time.     

Predicting the Outcome of Construction Litigation Using Boosted Decision Trees

Construction litigation has become commonplace in numerous construction projects, particularly in large contracts. Miscommunication, inadequate plans and specifications, rigid contracts, changes in site conditions, nonpayment, catch up profits, limitations on manpower, tools, and equipment, improper supervision, notice requirements, constructive changes not recognized as such by owner, delays, and acceleration measures provoke claims and often result in disputes. A boosted decision tree system was used to predict the outcome of construction litigation. The study was conducted by using the same 114 Illinois court cases that were used in earlier prediction studies conducted with artificial neural networks in 1998 and case-based reasoning in 1999, augmented by an additional 18 cases that were filed in 1990–2000. All cases were extracted from the Westlaw on-line service. The best prediction result obtained with boosted decision trees was 90%. The boosted decision tree model appears to be a promising tool to help create a dispute-free construction industry.     

Analysis of Masonry Panel under Biaxial Bending Using ANNs and CBR

In this paper the capability of artificial neural networks (ANNs) in solving complex nonlinear problems is utilized for the analysis of masonry panels under biaxial bending. A network, trained using a set of data, which is representative of the problem domain, is shown to be successful in solving new problems with reasonable accuracy. The experimental results obtained from the testing of panels are analyzed using the existing theories, and the method that gives good correlation between the theoretical prediction and the experimental result is recommended for other panels of similar properties and boundary conditions. An artificial intelligence based technology, the case-based reasoning (CBR), has been used to solve new problems by adapting solutions to similar problems solved in the past, which are stored in the case library. In this paper a hybrid system is described that utilizes the capabilities of both ANNs and CBR. CBR is used to identify a theoretical method that is most suitable for the present problem, whereas ANNs are used to arrive at a solution with great savings in computational time for the design of masonry panels subjected to biaxial bending.     

Design of Highway Bridges: Natural Place for CBR

The paper addresses a model, a framework, and an implemented system for supporting design activities where the use of case-based reasoning may reveal particular appropriateness. In the proposed environment, special attention is given to the synthesis of solutions by means of adaptation. A pragmatic combination of a number of artificial intelligence (AI) techniques, considering case-based reasoning (CBR) as the framing concept, enables the implementation of a system that conveniently supports most designers’ cognitive needs. The design of highway bridges was the chosen domain of discourse, for it represents an excellent example for demonstrating the potential of analogy in design. Thus, a large base of real cases is built. The induction of new knowledge is performed by extraction, association, and regression processes. Finally, a real context is used to illustrate the use of the model and to demonstrate its utility and capabilities in supporting designers’ decisions, particularly on the synthesis—i.e., adaptation—of solutions.     

磨矿过程设定值在线优化策略设计

针对基于传统模型的方法难以在线优化磨矿过程回路设定值的问题,提出了基于案例推理与强化学习的运行指标优化方法,建立基于自回归神经网络的Q函数模型,并应用案例推理更新模型连接权值,实现了磨矿过程关键参数的实时优化。     

The two stage case-based reasoning technology research on the basis of the expert decision

In some domain expert’s decision-making is intelligent output by comprehensive analysis and reasoning for the numerous decision-making factors,constraints and the goal in their domain,it is the concentrated expression of domain knowledge.Be aimed at field knowledge induction;sum up the exploration with systematization method,the two stages case-based reasoning(CBR) technology on the basis of the expert decision cased is present.The main idea,according to the characteristic of problem,analysis and decision process of field expert is carried out case-reasoning by two steps,and the field knowledge can be classified,induction and accumulated from different angles by multi case libraries.This technology includes two aspects,first,through two state case-based reasoning mechanisms to achieve area of decision-making process simulation;second,the CBR service in two more cases of library design process.First,two stage case-based reasoning mechanism,the first stage case-based reasoning to analyze the main level,clear nature of the problem,type,degree and character as the main content;the second stage case reasoning takes first order case reasoning result as basis then,the case reasoning,gives to making policy holding out being that purpose is in progress to out a decision-making suggestion once again.Secondly,in many cases the design of the library,to serve two-stage case-based reasoning process,the design of the four case libraries:First, property type case library,is a static case base can be modified,the effect is to performance differences between actual case and plan case of the qualitative distinction,indicate difference type attribute;Second, hierarchic quantization case library,is a static case base can be modified,the effect is to performance differences between actual case and plan case the quantitative distinction between the scope,indicate degree, size that the type issues in some difference etc.Three are that decision-making supports case library,is a static case base can be modified,the effect is similar enumeration all possible decision-makings content, indicate the decision-making that field expert possibility carries out;Four are decision match case library,is a self-study and dynamic case library,the effect is the library recording history decision-making case,is used to gain the decision-making scheme specifically for current reality achievement case is similar,to provide the support making policy,to learn composing in reply the case accumulating the new decision-making at the same time.The method is based on real business needs for research and development obtained after the conclusion,the paper shows through examples of the technology industry has a good adaptability,practicality and effectiveness.     

Preliminary Cost Estimation Model Using Case-Based Reasoning and Genetic Algorithms

This study proposes a preliminary cost estimation model using case-based reasoning (CBR) and genetic algorithm (GA). In measuring similarity and retrieving similar cases from a case base for minimum prediction error, it is a key process in determining the factors with the greatest weight among the attributes of cases in the case base. Previous approaches using experience, gradient search, fuzzy numbers, and analytic hierarchy process are limited in their provision of optimal solutions. This study therefore investigates a GA for weight generation and applies it to real project data. When compared to a conventional construction cost estimation model, the accuracy of the CBR- and GA-based construction cost estimation model was verified. It is expected that a more reliable construction cost estimation model could be designed in the early stages by using a weight estimation technique in the development of a construction cost estimation model.