RC-Net:Row and Column Network with Text Feature for Parsing Floor Plan Images

The popularity of online home design and floor plan customization has been steadily increasing. However, the manual conversion of floor plan images from books or paper materials into electronic resources can be a challenging task due to the vast amount of historical data available. By leveraging neural networks to identify and parse floor plans, the process of converting these images into electronic materials can be significantly streamlined. In this paper, we present a novel learning framework for automatically parsing floor plan images. Our key insight is that the room type text is very common and crucial in floor plan images as it identifies the important semantic information of the corresponding room. However, this clue is rarely considered in previous learning-based methods. In contrast, we propose the Row and Column network (RC-Net) for recognizing floor plan elements by integrating the text feature. Specifically, we add the text feature branch in the network to extract text features corresponding to the room type for the guidance of room type predictions. More importantly, we formulate the Row and Column constraint module (RC constraint module) to share and constrain features across the entire row and column of the feature maps to ensure that only one type is predicted in each room as much as possible, making the segmentation boundaries between different rooms more regular and cleaner. Extensive experiments on three benchmark datasets validate that our framework substantially outperforms other state-of-the-art approaches in terms of the metrics of FWIoU, mACC and mIoU.

Dynamic Neural Network Based Nonlinear Control of a Distillation Column

Taking advantage of the knowledge of top and bottom compositions of a distillation column, a dynamic neural network (DNN) is designed to identify the input-output relationship of the column. The weight-training algorithm is derived from a Lyapunov function. Based on this empirical model, a nonlinear H∞ controller is synthesized. The effectiveness of the control strategy is demonstrated using simulation results.

Control of liquid column height in electromagnetic casting with fuzzy neural network model

The control of suitable and stable height of liquid column is the crucial point to operate the electromagnetic casting(EMC) process and to obtain ingots with desirable shape and dimensional accuracy. But due to the complicated interact parameters and special circumstances, the measure and control of liquid column are quite difficult. A fuzzy neural network was used to help control the liquid column by predicting its height on line. The results show that the stabilization of the height of liquid column and surface quality of the ingot are remarkably improved by using the neural network based control system.

A Multi-Agent Immune Network Algorithm and Its Application to Murphree Efficiency Determination for the Distillation Column

Artificial Immune Network （aiNet） algorithms have become popular for global optimization in many modem industrial applications. However, high-dimensional systems using such models suffer from a potential premature convergence problem. In the existing aiNet algorithms, the premature convergence problem can be avoided by implementing various clonal selection methods, such as immune suppression and mutation approaches, both for single population and multi-population cases. This paper presents a new Multi-Agent Artificial Immune Network （Ma-aiNet） algorithm, which combines immune mechanics and multiagent technology, to overcome the premature convergence problem in high-dimensional systems and to efficiently use the agent ability of sensing and acting on the environment. Ma-aiNet integrates global and local search algorithms. The perform- ance of the proposed method is evaluated using 10 benchmark problems, and the results are compared with other well-known intelligent algorithms. The study demonstrates that Ma-aiNet outperforms other algorithms tested. Ma-aiNet is also used to determine the Murphree efficiency of a distillation column with satisfactory results.

Application of Artificial Neural Networks Based Monte Carlo Simulation in the Expert System Design and Control of Crude Oil Distillation Column of a Nigerian Refinery

This research work investigated comparative studies of expert system design and control of crude oil distillation column (CODC) using artificial neural networks based Monte Carlo (ANNBMC) simulation of random processes and artificial neural networks (ANN) model which were validated using experimental data obtained from functioning crude oil distillation column of Port-Harcourt Refinery, Nigeria by MATLAB computer program. Ninety percent (90%) of the experimental data sets were used for training while ten percent (10%) were used for testing the networks. The maximum relative errors between the experimental and calculated data obtained from the output variables of the neural network for CODC design were 1.98 error % and 0.57 error % when ANN only and ANNBMC were used respectively while their respective values for the maximum relative error were 0.346 error % and 0.124 error % when they were used for the controller prediction. Larger number of iteration steps of below 2500 and 5000 were required to achieve convergence of less than 10-7?for the training error using ANNBMC for both the design of the CODC and controller respectively while less than 400 and 700 iteration steps were needed to achieve convergence of 10-4?using ANN only. The linear regression analysis performed revealed the minimum and maximum prediction accuracies to be 80.65% and 98.79%;and 98.38% and 99.98% when ANN and ANNBMC were used for the CODC design respectively. Also, the minimum and maximum prediction accuracies were 92.83% and 99.34%;and 98.89% and 99.71% when ANN and ANNBMC were used for the CODC controller respectively as both methodologies have excellent predictions. Hence, artificial neural networks based Monte Carlo simulation is an effective and better tool for the design and control of crude oil distillation column.

Simulation and heat exchanger network designs for a novel single-column cryogenic air separation process

To realize the industrialization of the novel single-column air separation process proposed in previous work,steady-state simulation for four different configurations of the single-column process with ternary(nitrogen,oxygen and argon)is developed.Then,exergy analysis of the single-column processes is also carried out and compared with the conventional double-column air separation process at the same capacity.Furthermore,based on the steady-state simulation of single-column processes,the different heat exchanger networks(HENs)for the main heat exchanger and subcooler in each process are designed.To obtain better performance for this novel process,optimization of process configuration and operation is investigated.The optimal condition and configuration for this process is consisted as:feedstock is divided into two streams and the reflux nitrogen is compressed at the approximate temperature of 301 K.In addition,HEN is optimized to minimize the utilities.HENs without utilities are obtained for the four different configurations of single-column process.Furthermore,capital costs of the HEN for different cases are estimated and compared.

A Review of an Expert System Design for Crude Oil Distillation Column Using the Neural Networks Model and Process Optimization and Control Using Genetic Algorithm Framework

This paper presents a comprehensive review of various traditional systems of crude oil distillation column design, modeling, simulation, optimization and control methods. Artificial neural network (ANN), fuzzy logic (FL) and genetic algorithm (GA) framework were chosen as the best methodologies for design, optimization and control of crude oil distillation column. It was discovered that many past researchers used rigorous simulations which led to convergence problems that were time consuming. The use of dynamic mathematical models was also challenging as these models were also time dependent. The proposed methodologies use back-propagation algorithm to replace the convergence problem using error minimal method.

Recovery and grade prediction of pilot plant flotation column concentrate by a hybrid neural genetic algorithm

Today flotation column has become an acceptable means of froth flotation for a fairly broad range of applications, in particular the cleaning of sulfides. Even after having been used for several years in mineral processing plants, the full potential of the flotation column process is still not fully exploited. There is no prediction of process performance for the complete use of available control capabilities. The on-line estimation of grade usually requires a significant amount of work in maintenance and calibration of on-stream analyzers, in order to maintain good accuracy and high availability. These difficulties and the high cost of investment and maintenance of these devices have encouraged the approach of prediction of metal grade and recovery. In this paper, a new approach has been proposed for metallurgical performance prediction in flotation columns using Artificial Neural Network (ANN). Despite of the wide range of applications and flexibility of NNs, there is still no general framework or procedure through which the appropriate network for a specific task can be designed. Design and structural optimization of NNs is still strongly dependent upon the designer's experience. To mitigate this problem, a new method for the auto-design of NNs was used, based on Genetic Algorithm (GA). The new proposed method was evaluated by a case study in pilot plant flotation column at Sarcheshmeh copper plant. The chemical reagents dosage, froth height, air, wash water flow rates, gas holdup, Cu grade in the rougher feed, flotation column feed, column tail and final concentrate streams were used to the simulation by GANN. In this work, multi-layer NNs with Back Propagation (BP) algorithm with 8-17-10-2 and 8- 13-6-2 arrangements have been applied to predict the Cu and Mo grades and recoveries, respectively. The correlation coefficient (R) values for the testing sets for Cu and Mo grades were 0.93, 0.94 and for their recoveries were 0.93, 0.92, respectively. The results discussed in this paper indicate that the proposed model can be used to predict the Cu and Mo grades and recoveries with a reasonable error.

Semi-Active TLCD Control of Fixed Offshore Platforms Using Artifical Neural Networks

In this paper, the control method for fixed offshore platforms using semi-active tuned liquid column damper (TLCD) is presented. The equation of motion for the platform-TLCD control system is given and the semi-active control strategy is established. A back propagation artificial neural network (ANN) is used to adjust the orifice opening of TLCD because of the nonlinear motion of liquid in TLCD. The effectiveness of the control method is verified by numerical examples.

基于Column-and-constraint的鲁棒配电网重构实用算法

网络重构是配电自动化的重要功能之一,随着智能电网的迅速发展,也越发体现其重要性。但是负荷的波动性和随机性给传统的静态配电网重构带来了严峻的挑战,改进的两阶段鲁棒优化模型可以很好地解决配电网重构中存在的这一问题。第一阶段配置辐射状配电网络,第二阶段寻求满足需求的最优潮流。该策略采用Column-and-constraint实用算法,其主问题和子问题采用整数二阶锥规划算法(mixed-integer second-order cone programming algorithm)来解决。33节点、69节点等典型算例实验结果表明,即使在负荷不确定的情况下,鲁棒配电重构表现出比传统配电网络功率损耗低,可靠性高的特点。实验验证了方法的有效性和良好的计算性能。

Prediction of Axial Capacity of Concrete-Filled Square Steel Tubes Using Neural Networks

The application of artificial neural network to predict the ultimate bearing capacity of CFST （ concrete-filled square steel tubes） short columns under axial loading is explored. Input parameters consiste of concrete compressive strength, yield strength of steel tube, confinement index, sectional dimension and width-to-thickness ratio. The ultimate bearing capacity is the only output parameter. A multilayer feedforward neural network is used to describe the nonlinear relationships between the input and output variables. Fifty-five experimental data of CFST short columns under axial loading are used to train and test the neural network. A comparison between the neural network model and three parameter models shows that the neural network model possesses good accuracy and could be a practical method for predicting the ultimate strength of axially loaded CFST short columns.

基于ANN的RECFST短柱轴压承载力预测

目的针对相关设计规范和文献在计算圆端形截面钢管混凝土短柱轴压承载力上的局限性,开发高精高效的轴压承载力预测模型。方法首先,基于国内外已有的RECFST短柱轴压试验研究结果建立有限元模型,并通过验证;其次,基于Python脚本批量生成有限元模型,建立涵盖广泛输入参数的数据集;然后,利用数据集开发高精度的ANN模型并与相关规范和文献结果进行比较;最后,基于ANN模型开发GUI图形用户界面工具。结果ANN模型预测值与试验结果之比的平均值N ANN/N u=0.98,模型预测误差远低于相关规范和文献公式预测误差;ANN模型的均方误差K MSE=7.3734×10-7,总数据样本回归值R=0.99963,表明了ANN模型的有效性以及预测结果的精确性。结论ANN模型可以准确预测RECFST短柱的轴压承载力,基于模型开发的GUI工具简便实用。

基于SO-LSTM的立柱液压系统故障诊断方法研究

针对目前无法快速、准确地诊断矿用立柱液压系统故障等问题,在建立仿真模型分析单一故障机制的基础上,基于优化算法提出多种故障诊断方法。将立柱物理模块与立柱液压系统模块相结合,建立立柱液压系统仿真模型;基于Simulink分析单一故障的影响,基于蛇优化LSTM神经网络建立诊断模型;最后,根据实际数据进行模型的实例验证。结果表明:蛇优化LSTM模型对液压立柱故障仿真数据识别率达到99.5%,对液压立柱故障真实数据识别率达到97%,与模型仿真数据的预测精度仅相差2.5%,预测精度较高,达到了预期目标。

结合单列多列神经网络的移动状态人群计数方法研究

已有人群计数方法局限于对人群的全部进行计数,在仅对人群中的移动者进行计数时准确率较低,基于注意力的多阶段深度学习框架被提出以解决这一问题。通过注意力机制适应性地在单列和多列计数网络进行选择,结合单列网络的深层特征表示能力和多列网络多尺度特征学习能力,有效提取人群中移动者的特征。实验结果表明,所提出的方法均方误差(MSE)和平均绝对误差(MAE)皆低于已有人群计数方法,能够有效提高处于移动状态的人群的计数精度。

探析医院建筑柱网柱跨的合理确定

介绍了国内医院建筑柱网柱跨相关研究进展;分析了地下室停车位、“双通道”设计、未来开设公共服务空间的需求等因素对医院建筑柱网柱跨的影响;对8100 mm和8400 mm柱跨形成的诊室、病房进行了对比,并对柱跨选择提出了具体建议。

经验知识监督的RC墩柱力学性能神经网络分析方法

基于试验或数值模拟的单一墩柱力学性能分析方法难以兼顾计算精度和效率,纯数据驱动的分析方法存在可解释性差和对数据依赖性强等问题。为此,本文通过研究钢筋混凝土(RC)墩柱力学性能试验数据、经验知识和机器学习的融合机制,提出了经验知识监督的RC墩柱力学性能神经网络(knowledge-guided neural network,KGNN)分析方法。首先,建立了包含761组RC墩柱拟静力试验样本的数据库;随后,基于经验知识分析了RC墩柱主要特征对其力学性能的影响规律,构建了相应的数学表征方法;最后,将RC墩柱试验数据及经验知识融入人工神经网络架构和训练过程,建立了高精度、可解释、可通用且不依赖大量训练数据的RC墩柱力学性能KGNN分析模型。本文提出的KGNN分析方法与纯数据驱动神经网络(BPNN)的结果对比表明:BPNN在测试集上表现更好,在分析墩柱承载力时均方根误差(E)和拟合系数(R^(2))分别为0.070和0.978,KGNN模型的E和R^(2)分别为0.108和0.942;但由于BPNN所预测的墩柱特征对承载力的影响规律与经验知识并不吻合,即未能准确反映墩柱特征与其力学性能间的关系,BPNN模型发生了过拟合;而KGNN方法不仅可以快速准确获得RC墩柱力学性能,且预测规律与经验知识吻合较好,具有更高的可靠性和实用性。因此,融合试验数据与经验知识的神经网络有望成为一种新的RC结构力学性能分析方法。

电动汽车空间可调度特性影响下配电网承载能力计算方法

高渗透率分布式电源与电动汽车融合带来的波动性和随机性将对电网的安全稳定运行产生重大影响。如何准确计算配电网的承载能力已成为一个亟待研究和解决的问题。针对配电网承载能力的计算问题,提出了一种电动汽车空间可调度特性影响下配电网的承载能力分布鲁棒计算方法。首先,从电动汽车充电选择的角度出发,考虑车主自身利益、充电效率和配电网承载能力的提高,建立了电动汽车的空间可调度模型;然后,建立了两阶段分布鲁棒的分布式电源和充电站承载能力计算模型,不确定性概率分布的置信集同时受到1-范数和∞-范数的约束;最后,使用列和约束生成算法来解决该问题。通过相应的实例,分析了各种设备和EV空间可调度性对配电网承载能力的影响。通过与确定性模型和鲁棒模型的比较,验证了该模型的优越性。

工业过程关键指标预测的知识协同进化增强图卷积网络方法

在流程工业关键变量预测领域,已有研究致力于将过程知识与大数据相结合,以实现更高的准确性,降低过拟合风险和提高可解释性.然而,现有工作存在准确的先验知识构建成本高、无法从丰富的数据中挖掘知识等问题,限制了这些方法在实际工业过程中的广泛应用.为了解决这些挑战,本文提出了一种基于知识协同进化的增强图卷积网络方法.首先,利用易获取的过程流图构建低成本的粗粒度流程知识.然后,在图卷积神经网络模型训练中引入图探索,实现知识更新.最后,为了降低知识复杂度并保持一致性,设计了一种知识过滤机制.所提出的方法在基准的脱丁烷塔工艺过程上进行了验证.实验结果表明,该方法具有出色的预测准确性,并获得高质量的新知识.

基于LightGBM和DNNFL的陷落柱识别方法研究与应用

为了解决不平衡地震属性数据集中陷落柱识别准确率较低的问题,提出了一种基于轻梯度提升机(Light Gradient Boosting Machine,LightGBM)和利用焦点损失(Focal Loss)改进深度神经网络(Deep Neural Networks,DNN)相结合的陷落柱识别方法LightGBM-DNNF。首先通过相关性分析和重要性分析进行属性优选;其次提取LightGBM叶子节点的路径作为新的特征,并与原始数据集组合成新的数据集;最后输入到DNNFL模型中进行分类训练,预测地质构造类型。引入精确率(P)、召回率(R)、F1分数(F_(1-score))、曲线下面积(A_(UC))作为评价指标,基于3个矿区的数据集开展对比实验和消融实验。实验结果表明,与传统的机器学习和单一的集成学习算法相比,LightGBM-DNNFL模型的F_(1-score)和A_(UC)值均在93%以上,能有效识别陷落柱,且模型泛化能力更强。