• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>SPE 7th Latin American and Caribbean petroleum engineering conference >Gas Oil Color (ASTM) Inference with Neural Network in an Oil Refinery Distillation Column
【24h】

Gas Oil Color (ASTM) Inference with Neural Network in an Oil Refinery Distillation Column

机译：炼油厂蒸馏塔中具有神经网络的粗柴油颜色（ASTM）推断

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The first operation in an oil refinery is the atmosphericrndistillation. To maximize the extraction of some products likernGas Oil, a proper set of on line analysis instruments isrnrequired. These kinds of instruments are not always available,rnespecially in medium and small size processing plants.rnDue to the fact that color is a limiting specification, itrnconstitutes a restriction for production optimization.rnAvailability in real time of a good value estimate is whatrnallows work to be carried out permanently in operativernconditions where the process is most beneficial.rnIn this work a Neural Network (NN) approach to infer therncolor is proposed. A feed forward NN structure is used tornidentify the non-linear mapping from available processrnvariables to that property.rnTo acquire representative I/O data, a set of dynamicrnexperiments (move test) was developed in the plant. After that,rna rigorous analysis to select the set of input variables wasrnperformed. In this study, process engineer’s knowledge as wellrnas some mathematical tools were used to evaluate a minimumrnset of inputs. From this analysis, the set of forty-threernavailable inputs is reduced to the eight most sensitive withrnrespect to the color representation.rnFurthermore, rather than represent the entirerntransformation from the set of inputs variables to the outputrnvariable by a single neural network function, we analyze thernpossibility of breaking down the mapping into an initial pre-processingrnstage followed by a parameterized neural networkrnmodel.
机译：炼油厂的第一个操作是常压蒸馏。为了最大限度地提取某些产品（如天然气油），需要一套合适的在线分析仪器。这些类型的仪器并非总是可用，特别是在中小型加工厂中。rn由于颜色是一种限制规格，对生产优化构成了限制。rn实时提供高价值估算值是必须要做的在该过程最有利的可操作条件下永久执行。在这项工作中，提出了一种神经网络（NN）来推断色的方法。前馈NN结构用于识别从可用过程变量到该属性的非线性映射。为了获取代表性的I / O数据，在工厂中开发了一组动态实验（移动测试）。之后，进行了严格的分析以选择输入变量集。在这项研究中，过程工程师的知识以及一些数学工具被用来评估最小输入量。通过这种分析，将四十个可用输入的集合减少到八个最敏感的区域（相对于颜色表示）。此外，我们分析了可能性，而不是用单个神经网络函数来表示从输入变量到输出变量的整个变换。将映射分解为初始预处理阶段，然后是参数化神经网络模型。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号