WO2019178948A1 - 流程工业过程的一种多层模式监控方法 - Google Patents

流程工业过程的一种多层模式监控方法 Download PDF

Info

Publication number
WO2019178948A1
WO2019178948A1 PCT/CN2018/089589 CN2018089589W WO2019178948A1 WO 2019178948 A1 WO2019178948 A1 WO 2019178948A1 CN 2018089589 W CN2018089589 W CN 2018089589W WO 2019178948 A1 WO2019178948 A1 WO 2019178948A1
Authority
WO
WIPO (PCT)
Prior art keywords
mode
layer
monitoring method
mode monitoring
layer mode
Prior art date
Application number
PCT/CN2018/089589
Other languages
English (en)
French (fr)
Inventor
栾小丽
郑年年
冯恩波
刘成林
刘飞
Original Assignee
江南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江南大学 filed Critical 江南大学
Priority to US16/256,101 priority Critical patent/US11120350B2/en
Publication of WO2019178948A1 publication Critical patent/WO2019178948A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00Testing or monitoring of control systems or parts thereof
    • G05B23/02Electric testing or monitoring
    • G05B23/0205Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0259Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
    • G05B23/0267Fault communication, e.g. human machine interface [HMI]
    • G05B23/0272Presentation of monitored results, e.g. selection of status reports to be displayed; Filtering information to the user
    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00Programme-control systems
    • G05B19/02Programme-control systems electric
    • G05B19/418Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
    • G05B19/4185Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by the network communication
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063Operations research, analysis or management
    • G06Q10/0639Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00Program-control systems
    • G05B2219/30Nc systems
    • G05B2219/31From computer integrated manufacturing till monitoring
    • G05B2219/31088Network communication between supervisor and cell, machine group

Definitions

  • the invention relates to a multi-layer mode monitoring method for a process industrial process, belonging to the field of industrial production and processing.
  • a conventional parameter control method based on process measurement variables (such as temperature, pressure, liquid level, flow rate, composition, etc.) is to control one or more process parameters in a single factor to achieve a desired product quality or production. Operating indicators of the device.
  • process measurement variables such as temperature, pressure, liquid level, flow rate, composition, etc.
  • mode is essentially the representation of the similarity or difference of the data structure characteristics of multivariate complex systems.
  • the data vector of the observation space is projected to the space of the lower dimension with the minimum dimension and the maximum feature degree, and these lower orthogonal dimensions are called the principal elements.
  • the process parameter variable presents in the principal element feature space, which determines the efficiency of the device operation and the quality of the product. Therefore, the model can more clearly present the system essential feature states that are difficult to describe by the traditional process variable model method, such as normal state, abnormal state, early accident state, irreversible accident state, high yield state, low yield state, and the like.
  • the invention discloses a multi-layer mode monitoring method for a process industrial process, which divides an industrial process into multiple levels from a mode perspective, selects different key performance indicators for each level, and collects related key performance indicators.
  • the operation data is used to identify the modes of each level, and the mode monitoring methods of each level are proposed based on the data-driven method.
  • the multi-layer mode monitoring method includes:
  • Step 1 Perform multi-layer structure division of equipment devices in industrial processes
  • Step 2 Select key performance indicators of each layer in the multi-layer architecture obtained in step 1, and collect offline data related to key performance indicators of each layer in the multi-layer architecture;
  • Step 3 Perform cluster analysis method on the offline data collected in step 2, and identify patterns of each layer in the corresponding clustering plane respectively;
  • Step 4 Monitor real-time data according to the patterns of the layers identified in the clustering plane in step 3.
  • the method before the step of analyzing the collected offline data by using a cluster analysis method, the method further includes:
  • the mode in the step 3 includes at least one of a normal mode, a failure mode, an efficient mode, an intermediate efficiency and the like, an inefficient mode, and a desired mode.
  • the multi-layer mode monitoring method further includes:
  • the first predetermined manner includes a line graph mode, a bar graph mode, a histogram mode, and a scatter plot mode.
  • the multi-layer mode monitoring method further includes:
  • N is an integer greater than or equal to 2;
  • the display mode includes a horizontal histogram.
  • the multi-layer mode monitoring method further includes:
  • the spatial distance calculation method includes a Mahalanobis distance calculation method.
  • the multi-layer mode monitoring method further includes:
  • the probability of occurrence of the fault at the current moment is displayed in real time, and the fault handling method is given.
  • the multi-layer mode monitoring method further includes:
  • the remaining time of the fault occurrence is predicted and displayed in real time.
  • the multi-layer mode monitoring method further includes:
  • An alarm is issued when the predicted failure occurrence probability is greater than the first predetermined value and/or when the predicted remaining time of occurrence of the failure is less than the second predetermined value.
  • the cluster analysis method in the step 3 includes a principal component analysis, a K-means clustering method, a Bayesian classification method, and a potential function discriminant method.
  • performing the multi-layer architecture division of the equipment devices in the industrial process in the step 1 includes:
  • the equipment in the industrial process is divided into a device layer, an operation unit layer, a device layer, and a factory layer.
  • the device layer comprises one or more of a pump, a control valve, a pipeline, a heat exchanger, and a compressor.
  • the operating unit layer comprises one or more of a reactor, a heater, a rectification column, a shift furnace, a separator, a flasher, and an evaporator.
  • the device layer is a combination of at least two operating units; the factory layer is a combination of at least two devices.
  • the calculating the economic indicator value by combining the specific processes including:
  • the first predetermined manner includes a line graph mode, a bar graph mode, a histogram mode, and a scatter plot mode.
  • the alarm is performed when the remaining time of the predicted fault occurs is less than a predetermined value, and the alarm includes:
  • the message is pushed to the person in charge of the corresponding authority in a second predetermined manner, the second predetermined manner including at least one of a mail mode, a voice call mode, and a short message mode.
  • the multi-layer mode monitoring method further includes:
  • the cause of the failure and the corresponding processing suggestions are predicted and displayed; and the cause of the failure and the corresponding processing suggestions are saved to the accident database.
  • the invention provides a multi-layer mode monitoring method for a process industrial process, which realizes mode monitoring in a process industrial process, and compares with conventional parameter measurement based on process measurement variables.
  • the present invention simulates performance indexes of each layer as layers. In each mode, monitoring can be realized from micro to macro layers.
  • the real-time data is monitored by the patterns of each layer identified in the clustering plane to achieve the effect of quickly finding faults and affecting the current mode through selection.
  • the largest N variables combine the expert system and the inference engine to achieve the effect of troubleshooting, while reducing the energy consumption of the process, optimizing the operating costs and improving the competitiveness of the product.
  • FIG. 2 is a schematic diagram of a layered architecture of a process industrial process
  • Figure 3 is a schematic diagram of the principal element mode of the crude oil desalination and dehydration device layer
  • Figure 4 is an offline mode classification diagram of a crude oil desalination dehydration device
  • Figure 5 is a trajectory diagram of a past time period mode of a crude oil desalination dehydration device as a function of time;
  • Figure 6 is a graph showing the value loss of the crude oil desalination and dehydration device
  • Figure 7 is a horizontal histogram of the three most important variables of the crude oil desalination dehydration device sorted by the contribution.
  • a multi-layer mode monitoring method for the process industrial process divides the industrial process into multiple levels from a mode perspective, selects different key performance indicators for each level, and collects operational data related to key performance indicators, and identifies The mode of each level is proposed, and the mode monitoring method of each level is proposed based on the data driving method.
  • the multi-layer mode monitoring method includes:
  • Step 1 Perform multi-layer structure division of equipment devices in industrial processes
  • FIG. 2 is a schematic diagram of a hierarchical architecture of a process industrial process.
  • Step 2 Select key performance indicators of each layer in the multi-layer architecture obtained in step 1, and collect offline data related to key performance indicators of each layer in the multi-layer architecture;
  • the safe operation of the device is selected as a performance index, and 50 sets of offline data related to the performance index are collected;
  • Step 3 Perform cluster analysis method on the offline data collected in step 2, and identify patterns of each layer in the corresponding clustering plane respectively;
  • the operation unit layer for the device layer, the operation unit layer, the device layer, and the factory layer, on the basis of in-depth understanding of the process, respectively, in the clustering plane, draw the normal mode, the failure mode, the high efficiency mode, the medium efficiency mode of each layer, Inefficient mode, desired mode, etc., and displayed in the monitoring interface.
  • Step 4 Monitor real-time data according to the patterns of the layers identified in the clustering plane in step 3.
  • the real-time data is used to draw the pattern trajectory with time, and is displayed in real time in the monitoring interface.
  • the mode of real-time data at each moment in the past period of time can be known. Once it is found that it is in the failure mode, the person in charge can know that the current moment has failed.
  • the invention divides the device device in the industrial process into a multi-layer structure; selects key performance indicators of each layer in the divided multi-layer architecture, and collects offline data related to key performance indicators of each layer in the multi-layer architecture;
  • the clustered analysis method is used to analyze the collected offline data, and the patterns of each layer are respectively identified in the corresponding clustering plane; real-time data is monitored according to the patterns of the layers identified in the clustering plane, A function that monitors real-time data and can quickly detect faults.
  • Step 1 Perform multi-layer structure division of equipment devices in industrial processes
  • FIG. 2 is an example of a hierarchical architecture of a process industrial process.
  • Step 2 Select the key performance indicators of each layer in the multi-layer architecture obtained in step one, and collect the offline data related to the key performance indicators of each layer in the multi-layer architecture and eliminate the singular points;
  • Figure 3 is a schematic diagram of the principal element mode of a crude oil desalination dehydration unit. The difference between the set of data on the right side of Figure 3 and the overall data pattern is too large, and this set of data should be eliminated.
  • Step 3 The offline data collected in step 2 is analyzed by using a cluster analysis method, and the modes of each layer are respectively identified in the corresponding clustering plane;
  • the analysis by using the principal component analysis method is taken as an example.
  • any one of the K-means clustering method, the Bayesian classification method, and the potential function discriminant method may be used.
  • Cluster analysis any one of the K-means clustering method, the Bayesian classification method, and the potential function discriminant method may be used.
  • Step 4 monitor real-time data according to the patterns of the layers identified in the clustering plane in step 3;
  • FIG. 5 is a schematic diagram of the past time period mode of the crude oil desalination and dehydration device changing with time.
  • Step 5 Calculate the distance between the projection point of the current time mode on the clustering plane and the desired mode according to the spatial distance calculation method; the spatial distance calculation method includes the Mahalanobis distance calculation method.
  • the Mahalanobis distance calculation method is taken as an example to calculate the distance between the projection point of the mode at the current time and the desired mode in the two principal planes.
  • Step 6 According to the distance between the projection point on the clustering plane and the desired mode of the current moment mode calculated in step 5, combined with the crude oil desalination and dehydration process, the calculated distance is converted into a specific economic value, and the graph is The method displays the lost value in real time.
  • real-time display can be performed by any one of a line graph method, a bar graph method, a histogram method, and a scatter graph method.
  • Fig. 6 is a graph showing the value loss of the crude oil desalination dehydration device.
  • Step 7 Select N variables having the greatest influence on the current mode corresponding to the real-time data, where N is an integer greater than or equal to 2; sequentially display the influence degree of each variable on the current mode, and the display manner includes a horizontal histogram;
  • the main element reconstruction method such as the contribution graph method is taken as an example to find the three most important variables affecting the current time pattern, sorted according to the contribution degree of each variable, and contributed by the horizontal histogram. Sort the degree and display it in real time in the monitoring interface.
  • Figure 7 is a horizontal histogram of the three most important variables of the crude oil desalination dehydration unit sorted by the degree of contribution.
  • Step 8 Calculate the distance between the projection point of the current time mode on the clustering plane and the fault mode according to the spatial distance calculation method; the spatial distance calculation method includes the Mahalanobis distance calculation method.
  • the Mahalanobis distance calculation method is taken as an example to calculate the distance between the projection point of the mode of the crude oil desalination and dehydration device on the two principal planes and the failure mode.
  • Step 9 Calculate the probability of occurrence of the fault at the current moment according to the calculated distance between the projection point on the clustering plane and the fault mode of the current moment mode; display the fault occurrence probability at the current moment in real time;
  • the probability of occurrence of the failure of the crude oil desalination dehydration device is calculated, and the probability of occurrence of the current moment fault is displayed in real time in the monitoring interface.
  • Step 10 When the predicted fault occurs greater than the first predetermined value, an alarm is performed, and according to the selected N variables having the greatest influence on the current mode, combined with the expert system and the inference engine, the fault cause and corresponding processing suggestions are predicted and displayed. And save the cause of the failure and the corresponding processing suggestions to the accident database;
  • Step 11 predicting the remaining time of the fault occurrence according to the calculated distance and displaying it in real time; displaying the remaining time of the fault in real time and performing an alarm when the remaining time of the predicted fault occurs is less than the second predetermined value;
  • step 8 According to the distance between the projection point on the clustering plane and the fault mode of the mode at the current time calculated in step 8, predict the remaining time that the fault is about to occur, and if the remaining time is lower than a certain time (such as 1 hour), an alarm is issued. And show the remaining time.
  • a certain time such as 1 hour
  • step 10 and step 11 above when the predicted fault occurs greater than the first predetermined value and/or when the predicted remaining fault occurs less than the second predetermined value, the alarm is performed, ie, either If a situation occurs, an alarm is generated, and the message is pushed to the person in charge of the corresponding authority in a second predetermined manner, where the second predetermined manner includes at least one of a mail mode, a voice call mode, and a short message mode; in an actual application, The message may be pushed to the person in charge of the corresponding authority by any means, or may be pushed to the person in charge of the corresponding authority in multiple ways at the same time.
  • the first predetermined value may be set to 90% in the step 10 according to the actual application, or may be other values; the second predetermined value may be set to 1 hour in the step 11 according to the actual application, or may be combined with other actual values. .
  • the invention divides the device device in the industrial process into a multi-layer structure; selects key performance indicators of each layer in the divided multi-layer architecture, and collects offline data related to key performance indicators of each layer in the multi-layer architecture;
  • the clustered analysis method is used to analyze the collected offline data, and the patterns of each layer are respectively identified in the corresponding clustering plane; real-time data is monitored according to the patterns of the layers identified in the clustering plane, The function of monitoring real-time data and quickly detecting faults; also alarming by calculating the probability of occurrence of the fault and the remaining time of the fault, by using the N variables that have the greatest influence on the current mode corresponding to the real-time data and matching the current variables according to the respective variables.
  • the mode influence degree is displayed in sequence, and the expert system and the inference engine are combined to predict the cause of the fault and the corresponding processing suggestions, so that the fault can be eliminated in time after the fault is found, and the fault cause and the corresponding processing suggestion are saved to the accident database. In order to resolve similar faults later, you can quickly troubleshoot.
  • the above monitoring method achieves the effect of reducing the energy consumption of the process, optimizing the operating cost and improving the competitiveness of the product.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Educational Administration (AREA)
  • Automation & Control Theory (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Game Theory and Decision Science (AREA)
  • Human Computer Interaction (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

流程工业过程的一种多层模式监控方法,属于工业生产加工领域。通过从模式的角度将工业过程划分为多个层级,针对各个层级选取不同的关键性能指标,采集与关键性能指标相关的运行数据,标识出各个层级的模式,基于数据驱动方法提出各个层级的模式监控方法,实现了流程工业过程中模式监控。通过聚类平面内标识出的各层的模式对实时数据进行监控实现了快速发现故障的效果,并通过选取的对当前模式影响最大的N个变量结合专家系统和推理机实现排除故障的效果,同时降低了过程的能源消耗,优化了运行成本并提高了产品的竞争力。

Description

流程工业过程的一种多层模式监控方法 技术领域
本发明涉及流程工业过程的一种多层模式监控方法,属于工业生产加工领域。
背景技术
过程安全、产品质量以及节能减排增效是现代流程工业的核心目标,因此对过程的运行状态进行实时监控具有重大的现实意义和价值。随着现代流程工业过程装置规模的扩张,工艺复杂度的上升,过程运行的安全性及市场对产品个性化的要求等,使得对复杂流程工业的过程运行进行状态监控更加迫切。常规基于过程测量变量(如温度,压力,液位,流量,成分等)的参数控制方法是将一个或多个过程参数以单因素方式分别控制在一定范围内,从而达到期望的产品品质或者生产装置的运行指标。但是工业实践中,即便将所有的过程参数都控制在各自的期望范围内,产品最终品质的一致性有时不能保证,甚至还不一定满足生产要求。
究其原因,实际生产过程的参数变量具有多因素特性,使得过程参数变量在整体上呈现着模式行为。所谓模式,本质上是多变量复杂系统数据结构特征的相似性或差异性的表现。数学上可以表示为,观测空间的数据向量以最小维数和最大特征区分度投影到较低维数的空间,这些较低的正交维度被称为主元。正是过程参数变量在主元特征空间呈现出的“模式”,决定了装置运行的效率和产品的品质。因此,模式能够更清楚地呈现出传统过程变量模型方法难以刻画的系统本质特征状态,如正常状态、非正常状态、事故早期状态、不可逆转事故状态、高产率状态、低产率状态等等。
同时随着计算机存储技术的飞速发展,海量过程数据得以收集和存储。但由于缺少充足的过程经验和应用工具,工业过程往往数据丰富但知识缺乏。现有的数据监控仅仅限于监控过程参数,并不能从宏观上、多层次的展现流程工业中的生产过程,使得在发现问题、预估损失、采取补救措施的时间较长,不能及时止损而造成更为严重的经济和能源的损失。因此,提供流程工业过程的一种多层模式监控方法对于丰富和发展过程控制理论具有重要意义,同时对于降低过程的能源消耗,优化运行成本以及提高产品的竞争力等均有重要的理论价值和现实意义,在流程工业中定有广泛的应用前景。
发明内容
本发明公开了流程工业过程的一种多层模式监控方法,所述方法是从模式的角度将工业过程划分为多个层级,针对各个层级选取不同的关键性能指标,采集与关键性能指标相关的 运行数据,标识出各个层级的模式,基于数据驱动方法提出各个层级的模式监控方法。
可选的,所述多层模式监控方法包括:
步骤1:对工业过程中的设备装置进行多层架构划分;
步骤2:选取步骤1中划分得到的多层架构中每一层的关键性能指标,并采集多层架构中每一层的与关键性能指标相关的离线数据;
步骤3:对步骤2中采集到的离线数据采用聚类分析方法进行分析,并分别在对应的聚类平面中标识出各层的模式;
步骤4:根据步骤3中聚类平面内标识出的各层的模式对实时数据进行监控。
可选的,所述步骤3中对采集到的离线数据采用聚类分析方法进行分析之前,还包括:
剔除奇异点。
可选的,所述步骤3中的模式包括正常模式、故障模式、高效模式、中效等效率模式、低效模式、期望模式中的至少一种
可选的,所述多层模式监控方法还包括:
结合具体工艺计算经济指标数值,并以第一预定方式实时显示所述经济指标数值与实时数据的关系;
所述第一预定方式包括折线图方式、条形图方式,柱状图方式和散点图方式。
可选的,所述多层模式监控方法还包括:
选取对实时数据所对应的当前模式影响最大的N个变量,其中N为大于等于2的整数;
按照各个变量对当前模式的影响度大小进行顺序显示,显示方式包括横向柱状图。
可选的,所述多层模式监控方法还包括:
根据空间距离计算方法,计算当前时刻的模式在聚类平面上的投影点与故障模式的距离;
根据计算出的距离计算当前时刻的故障发生概率;
所述空间距离计算方法包括马氏距离计算法。
可选的,所述多层模式监控方法还包括:
实时显示当前时刻的故障发生概率,并给出故障的处理方法。
可选的,所述多层模式监控方法还包括:
根据计算出的距离,预测故障发生的剩余时间并实时显示。
可选的,所述多层模式监控方法还包括:
当预测出的故障发生概率的大于第一预定值和/或当预测出的故障发生的剩余时间小于第二预定值时进行报警。
可选的,所述步骤3中的聚类分析方法包括主元分析、K均值聚类方法、贝叶斯分类方法和势函数判别法。
可选的,所述步骤1中对工业过程中的设备装置进行多层架构划分包括:
将工业过程中的设备装置划分为设备层、操作单元层、装置层和工厂层。
可选的,所述设备层包括泵、控制阀、管道、换热器、压缩机中的一种或多种。
可选的,所述操作单元层包括反应器、加热器、精馏塔、变换炉、分离器、闪蒸器、蒸发器中的一种或多种。
可选的,所述装置层为至少两个操作单元的组合;所述工厂层为至少两个装置的组合。
可选的,所述结合具体工艺计算经济指标数值,包括:
根据空间距离计算方法,计算当前时刻的模式在聚类平面上的投影点与期望模式的距离;
结合具体工艺将计算出的距离转化为经济指标数值;
以第一预定方式将计算出的经济指标数值实时显示;
所述第一预定方式包括折线图方式、条形图方式,柱状图方式和散点图方式。
可选的,所述当预测出的故障发生的剩余时间小于预定值时进行报警中,所述报警包括:
以第二预定方式向相应权限的负责人推送消息,所述第二预定方式包括邮件方式、语音电话方式和短消息方式中的至少一种。
可选的,所述多层模式监控方法还包括:
根据所选取的对当前模式影响最大的N个变量,结合专家系统和推理机,预测并显示故障原因和对应的处理建议;并将故障原因和对应的处理建议保存至事故数据库。
本发明的有益效果:
本发明提供了流程工业过程的一种多层模式监控方法,实现了流程工业过程中的模式监控,相对与常规基于过程测量变量的参数监控,本发明是将各层的性能指标模拟为各层中的各个模式中,从微观到宏观各个层都能实现监控,通过聚类平面内标识出的各层的模式对实时数据进行监控实现了快速发现故障的效果,并通过选取的对当前模式影响最大的N个变量结合专家系统和推理机实现排除故障的效果,同时降低了过程的能源消耗,优化了运行成本并提高了产品的竞争力。
附图说明
图1为本发明实施例实施步骤流程图;
图2为一种流程工业过程的分层构架示意图;
图3为原油脱盐脱水装置层主元模式图;
图4为原油脱盐脱水装置的离线模式分类图;
图5为原油脱盐脱水装置随时间变化的过去时间段模式轨迹图;
图6为原油脱盐脱水装置价值流失曲线图;
图7为原油脱盐脱水装置最重要的三个变量按贡献度大小排序的横向柱状图。
具体实施方式
下述实施例以原油加工过程为例对本发明的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。
实施例一
本实施提供的流程工业过程的一种多层模式监控方法,从模式的角度将工业过程划分为多个层级,针对各个层级选取不同的关键性能指标,采集与关键性能指标相关的运行数据,标识出各个层级的模式,基于数据驱动方法提出各个层级的模式监控方法,参见图1,所述多层模式监控方法包括:
步骤1:对工业过程中的设备装置进行多层架构划分;
具体的,对原油加工的生产过程的设备装置进行划分,泵、控制阀、管道,换热器等划分为设备层,反应器、加热器、精馏塔等划分为操作单元级,数个操作单元组合为装置层,数个装置组成工厂层。图2是一种流程工业过程的分层构架示意图。
步骤2:选取步骤1中划分得到的多层架构中每一层的关键性能指标,并采集多层架构中每一层的与关键性能指标相关的离线数据;
以原油脱盐脱水装置层为例,选取装置的安全操作为性能指标,收集与该性能指标相关的50组离线数据;
步骤3:对步骤2中采集到的离线数据采用聚类分析方法进行分析,并分别在对应的聚类平面中标识出各层的模式;
具体的,针对设备层、操作单元层、装置层、工厂层,在深入理解工艺的基础上,分别在聚类平面中,画出各层的正常模式、故障模式、高效模式、中等效率模式、低效模式、期望模式等,并在监控界面中进行显示。
步骤4:根据步骤3中聚类平面内标识出的各层的模式对实时数据进行监控。
针对设备层、操作单元层、装置层、工厂层,在各层的历史模式聚类平面中,分别利用实时数据画出随着时间变化的模式轨迹,并在监控界面中实时显示。
根据实时数据随时间变化的过去时间段模式轨迹图即可获知实时数据在过去一段时间内每个时刻所处的模式,一旦发现其处于故障模式中,负责人即可获知当前时刻发生了故障。
本发明通过对工业过程中的设备装置进行多层架构划分;选取划分得到的多层架构中每一层的关键性能指标,并采集多层架构中每一层的与关键性能指标相关的离线数据;对采集到的离线数据采用聚类分析方法进行分析,并分别在对应的聚类平面中标识出各层的模式;根据聚类平面内标识出的各层的模式对实时数据进行监控,实现了对实时数据进行监控并能够快速发现故障的功能。
实施例二
本发明具体实施步骤和算法如下:
步骤一:对工业过程中的设备装置进行多层架构划分;
具体的,以原油加工流程工业过程为例,对生产过程的设备装置进行划分,其中,泵、控制阀、管道,换热器等划分为设备层;反应器、加热器、精馏塔等划分为操作单元级;数个操作单元组合为装置层;数个装置组成工厂层;图2是一种流程工业过程的分层构架示例。
步骤二:选取步骤一中划分得到的多层架构中每一层的关键性能指标,并采集多层架构中每一层的与关键性能指标相关的离线数据并剔除奇异点;
针对不同的设备层、操作单元层、装置层、工厂层,分别选取不同的关键性能指标,采集与关键性能指标相关的运行数据,进行聚类分析并在聚类平面中剔除奇异点,使得所有数据点在一个可信度之内。
以原油脱盐脱水装置层为例,选取装置的安全操作为性能指标,收集与指标相关的50组运行数据,进行聚类分析,这里以采用主元分析法进行分析为例进行说明。图3是一个原油脱盐脱水装置层主元模式图,图3右边有一组数据与整体数据模式差距太大,应剔除这一组数据。
步骤三:对步骤二中采集到的离线数据采用聚类分析方法进行分析,并分别在对应的聚类平面中标识出各层的模式;
具体的,本实施例中以采用主元分析法进行分析为例进行说明,实际应用中,还可以采用K均值聚类方法、贝叶斯分类方法和势函数判别法中的任意一种方法进行聚类分析;
针对设备层、操作单元层、装置层、工厂层,在深入理解工艺的基础上,分别在主元平面中,画出各层的正常模式、故障模式、高效模式、中等效率模式、低效模式、期望模式等,并在监控界面中进行显示。以步骤二中的原油脱盐脱水装置层为例,利用剩余的49组数据, 结合原油脱盐脱水工工艺,在两主元平面中离线标识出原油脱盐脱水装置的正常模式、故障模式、高效模式、中等效率模式、低效模式、期望模式。图4是原油脱盐脱水装置的离线模式分类图。
步骤四:根据步骤三中聚类平面内标识出的各层的模式对实时数据进行监控;
针对设备层、操作单元层、装置层、工厂层,在各层的历史模式主元平面中,分别利用实时数据画出随着时间变化的模式轨迹,并在监控界面中实时显示。以步骤二中的原油脱盐脱水装置层为例,图5是原油脱盐脱水装置随时间变化的过去时间段模式轨迹图。
步骤五:根据空间距离计算方法,计算当前时刻的模式在聚类平面上的投影点与期望模式的距离;空间距离计算方法包括马氏距离计算法。
这里以马氏距离计算法为例进行说明,计算当前时刻的模式在两主元平面中的投影点与期望模式的距离。
步骤六:根据步骤五中计算出的当前时刻的模式在聚类平面上的投影点与期望模式的距离,结合原油脱盐脱水工艺,将计算出的距离转化为具体经济数值,并以曲线图的方式将流失价值进行实时显示,实际应用中,可以采用折线图方式、条形图方式,柱状图方式和散点图方式中的任意一种方式进行实时显示。
如图6所示,图6是原油脱盐脱水装置价值流失曲线图。
步骤七:选取对实时数据所对应的当前模式影响最大的N个变量,其中N为大于等于2的整数;按照各个变量对当前模式的影响度大小进行顺序显示,显示方式包括横向柱状图;
本实施例这以通过主元重构法如贡献图法为例,找出影响当前时刻模式最重要的三个变量,按照各变量的贡献度大小进行排序,并以横向柱状图的方式按贡献度大小排序,在监控界面中实时显示。
图7是原油脱盐脱水装置最重要的三个变量按贡献度大小排序的横向柱状图。
步骤八:根据空间距离计算方法,计算当前时刻的模式在聚类平面上的投影点与故障模式的距离;空间距离计算方法包括马氏距离计算法。
这里以马氏距离计算法为例进行说明,计算原油脱盐脱水装置当前时刻的模式在两主元平面上的投影点与故障模式的距离。
步骤九:根据计算出的当前时刻的模式在聚类平面上的投影点与故障模式的距离计算当前时刻的故障发生概率;实时显示当前时刻的故障发生概率;
根据步骤八中计算出的当前时刻的模式在聚类平面上的投影点与故障模式的距离,计算原油脱盐脱水装置故障发生的概率,并在监控界面中实时显示当前时刻故障发生概率的大小。
步骤十:当预测出的故障发生的大于第一预定值时进行报警,根据所选取的对当前模式影响最大的N个变量,结合专家系统和推理机,预测并显示故障原因和对应的处理建议;并将故障原因和对应的处理建议保存至事故数据库;
具体的,若当前时刻的故障概率超过一定阈值时(如90%),则进行报警。同时利用专家系统和推理机,结合步骤七中重构出的三个变量,给出故障原因和处理建议,并将建议在监控界面中进行显示,同时保存到事故数据库进行存储。
步骤十一:根据计算出的距离,预测故障发生的剩余时间并实时显示;当预测出的故障发生的剩余时间小于第二预定值时实时显示故障发生剩余时间并进行报警;
根据步骤八中计算出的当前时刻的模式在聚类平面上的投影点与故障模式的距离,预测故障即将发生的剩余时间,若剩余时间低于特定时间(如1个小时),则进行报警并显示剩余时间。
具体的,上述步骤十和步骤十一中,当预测出的故障发生的大于第一预定值和/或当预测出的故障发生的剩余时间小于第二预定值时进行报警,即二者中任意一种情况发生则进行报警,并以第二预定方式向相应权限的负责人推送消息,所述第二预定方式包括邮件方式、语音电话方式和短消息方式中的至少一种;实际应用中,可采用任意一种方式向相应权限的负责人推送消息,也可以同时采用多种方式向相应权限的负责人推送消息。第一预定值可以根据实际应用设置为步骤十中的90%,也可以是其他数值;第二预定值可以根据实际应用设置为步骤十一中的1个小时,也可以是结合实际的其他数值。
本发明通过对工业过程中的设备装置进行多层架构划分;选取划分得到的多层架构中每一层的关键性能指标,并采集多层架构中每一层的与关键性能指标相关的离线数据;对采集到的离线数据采用聚类分析方法进行分析,并分别在对应的聚类平面中标识出各层的模式;根据聚类平面内标识出的各层的模式对实时数据进行监控,实现了对实时数据进行监控并能够快速发现故障的功能;还通过计算故障发生概率以及故障发生的剩余时间进行报警,通过对实时数据所对应的当前模式影响最大的N个变量并按照各个变量对当前模式的影响度大小进行顺序显示,还结合专家系统和推理机,预测故障原因和对应的处理建议,使得在发现故障后可以及时排除故障,通过将故障原因和对应的处理建议保存至事故数据库,以便后期再发生类似故障时可以快速排除故障。综合以上监控方法达到了降低过程的能源消耗,优化运行成本以及提高产品的竞争力的效果。
虽然本发明已以较佳实施例公开如上,但其并非用以限定本发明,任何熟悉此技术的人, 在不脱离本发明的精神和范围内,都可做各种的改动与修饰,因此本发明的保护范围应该以权利要求书所界定的为准。

Claims (18)

  1. 流程工业过程的一种多层模式监控方法,其特征在于,所述方法是从模式的角度将工业过程划分为多个层级,针对各个层级选取不同的关键性能指标,采集与关键性能指标相关的运行数据,标识出各个层级的模式,基于数据驱动方法提出各个层级的模式监控方法。
  2. 根据权利要求1所述的多层模式监控方法,其特征在于,所述多层模式监控方法包括:
    步骤1:对工业过程中的设备装置进行多层架构划分;
    步骤2:选取步骤1中划分得到的多层架构中每一层的关键性能指标,并采集多层架构中每一层的与关键性能指标相关的离线数据;
    步骤3:对步骤2中采集到的离线数据采用聚类分析方法进行分析,并分别在对应的聚类平面中标识出各层的模式;
    步骤4:根据步骤3中聚类平面内标识出的各层的模式对实时数据进行监控。
  3. 根据权利要求2所述的多层模式监控方法,其特征在于,所述步骤3中对采集到的离线数据采用聚类分析方法进行分析之前,还包括:
    剔除奇异点。
  4. 根据权利要求2所述的多层模式监控方法,其特征在于,所述步骤3中的模式包括正常模式、故障模式、高效模式、中效等效率模式、低效模式、期望模式中的至少一种
  5. 根据权利要求2所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包括:
    结合具体工艺计算经济指标数值,并以第一预定方式实时显示所述经济指标数值与实时数据的关系;
    所述第一预定方式包括折线图方式、条形图方式,柱状图方式和散点图方式。
  6. 根据权利要求2所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包括:
    选取对实时数据所对应的当前模式影响最大的N个变量,其中N为大于等于2的整数;
    按照各个变量对当前模式的影响度大小进行顺序显示,显示方式包括横向柱状图。
  7. 根据权利要求2所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包 括:
    根据空间距离计算方法,计算当前时刻的模式在聚类平面上的投影点与故障模式的距离;
    根据计算出的距离计算当前时刻的故障发生概率;
    所述空间距离计算方法包括马氏距离计算法。
  8. 根据权利要求7所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包括:
    实时显示当前时刻的故障发生概率,并给出故障的处理方法。
  9. 根据权利要求7所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包括:
    根据计算出的距离,预测故障发生的剩余时间并实时显示。
  10. 根据权利要求8或9所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包括:
    当预测出的故障发生概率的大于第一预定值和/或当预测出的故障发生的剩余时间小于第二预定值时进行报警。
  11. 根据权利要求2所述的多层模式监控方法,其特征在于,所述步骤3中的聚类分析方法包括主元分析、K均值聚类方法、贝叶斯分类方法和势函数判别法。
  12. 根据权利要求2所述的多层模式监控方法,其特征在于,所述步骤1中对工业过程中的设备装置进行多层架构划分包括:
    将工业过程中的设备装置划分为设备层、操作单元层、装置层和工厂层。
  13. 根据权利要求12所述的多层模式监控方法,其特征在于,所述设备层包括泵、控制阀、管道、换热器、压缩机中的一种或多种。
  14. 根据权利要求12或13所述多层模式监控方法,其特征在于,所述操作单元层包括反应器、加热器、精馏塔、变换炉、分离器、闪蒸器、蒸发器中的一种或多种。
  15. 根据权利要求12~14任一所述的多层模式监控方法,其特征在于,所述装置层为至少两个操作单元的组合;所述工厂层为至少两个装置的组合。
  16. 根据权利要求5所述的多层模式监控方法,其特征在于,所述结合具体工艺计算经济指标数值,包括:
    根据空间距离计算方法,计算当前时刻的模式在聚类平面上的投影点与期望模式的距离;
    结合具体工艺将计算出的距离转化为经济指标数值;
    以第一预定方式将计算出的经济指标数值实时显示;
    所述第一预定方式包括折线图方式、条形图方式,柱状图方式和散点图方式。
  17. 根据权利要求10所述的多层模式监控方法,其特征在于,所述当预测出的故障发生的剩余时间小于预定值时进行报警中,所述报警包括:
    以第二预定方式向相应权限的负责人推送消息,所述第二预定方式包括邮件方式、语音电话方式和短消息方式中的至少一种。
  18. 根据权利要求6所述的多层模式监控方法,其特征在于,所述多层模式监控方法还包括:
    根据所选取的对当前模式影响最大的N个变量,结合专家系统和推理机,预测并显示故障原因和对应的处理建议;并将故障原因和对应的处理建议保存至事故数据库。
PCT/CN2018/089589 2018-03-21 2018-06-01 流程工业过程的一种多层模式监控方法 WO2019178948A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/256,101 US11120350B2 (en) 2018-03-21 2019-01-24 Multilevel pattern monitoring method for industry processes

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810232872.9 2018-03-21
CN201810232872.9A CN108388221B (zh) 2018-03-21 2018-03-21 流程工业过程的一种多层模式监控方法

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/256,101 Continuation US11120350B2 (en) 2018-03-21 2019-01-24 Multilevel pattern monitoring method for industry processes

Publications (1)

Publication Number Publication Date
WO2019178948A1 true WO2019178948A1 (zh) 2019-09-26

Family

ID=63068218

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/089589 WO2019178948A1 (zh) 2018-03-21 2018-06-01 流程工业过程的一种多层模式监控方法

Country Status (2)

Country Link
CN (1) CN108388221B (zh)
WO (1) WO2019178948A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104238545A (zh) * 2014-07-10 2014-12-24 中国石油大学(北京) 一种石油炼化生产过程故障诊断和预警系统及其建立方法
CN104298187A (zh) * 2014-06-12 2015-01-21 东北大学 金湿法冶金全流程三层结构过程监测方法
CN105893700A (zh) * 2016-04-26 2016-08-24 陆新建 基于物理-大数据混合模型的化工生产在线故障检测与诊断技术
CN106054832A (zh) * 2016-06-07 2016-10-26 蓝星(北京)技术中心有限公司 基于多变量的间歇化工生产过程动态在线监测方法与装置

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101169623B (zh) * 2007-11-22 2010-07-07 东北大学 基于核主元分析贡献图的非线性过程故障辨识方法
CN101446827B (zh) * 2008-11-06 2011-06-22 西安交通大学 一种流程工业系统的过程故障分析装置及方法
US9378256B2 (en) * 2013-11-15 2016-06-28 Ut-Battelle, Llc Industrial geospatial analysis tool for energy evaluation
CN103617105B (zh) * 2013-12-17 2016-08-17 中国科学院合肥物质科学研究院 一种基于数据驱动的自适应多层流模型设备诊断方法
CN104573850B (zh) * 2014-12-18 2017-08-11 湖南大唐先一科技有限公司 一种火电厂设备状态评估方法
CN104504271A (zh) * 2014-12-25 2015-04-08 重庆科技学院 基于独立元相似度的高含硫天然气净化脱硫过程检测方法
US10509396B2 (en) * 2016-06-09 2019-12-17 Rockwell Automation Technologies, Inc. Scalable analytics architecture for automation control systems
CN106324405A (zh) * 2016-09-07 2017-01-11 南京工程学院 一种基于改进主成分分析的变压器故障诊断方法
CN106371427B (zh) * 2016-10-28 2019-03-29 浙江大学 基于层次分析法和模糊融合的工业过程故障分类方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104298187A (zh) * 2014-06-12 2015-01-21 东北大学 金湿法冶金全流程三层结构过程监测方法
CN104238545A (zh) * 2014-07-10 2014-12-24 中国石油大学(北京) 一种石油炼化生产过程故障诊断和预警系统及其建立方法
CN105893700A (zh) * 2016-04-26 2016-08-24 陆新建 基于物理-大数据混合模型的化工生产在线故障检测与诊断技术
CN106054832A (zh) * 2016-06-07 2016-10-26 蓝星(北京)技术中心有限公司 基于多变量的间歇化工生产过程动态在线监测方法与装置

Also Published As

Publication number Publication date
CN108388221B (zh) 2019-06-21
CN108388221A (zh) 2018-08-10

Similar Documents

Publication Publication Date Title
Ji et al. Incipient fault detection with smoothing techniques in statistical process monitoring
US11120350B2 (en) Multilevel pattern monitoring method for industry processes
JP5179086B2 (ja) 工業プロセスの監視方法及び監視システム
CN102252843B (zh) 一种滚动轴承性能变异的评估方法
CN104267668A (zh) 基于贝叶斯方法的航天阀门零件加工过程故障诊断方法
CN112529320A (zh) 空压机集群智能诊断系统
CN107862324B (zh) 一种基于mwspca的cbr预测模型智能化预警方法
CN110209144B (zh) 基于动静协同差异分析的两层实时监测与报警溯源方法
CN112306036B (zh) 化工过程运行故障的诊断方法
CN112861350B (zh) 一种水冷式汽轮发电机定子绕组温度过热缺陷预警方法
CN112598144B (zh) 基于相关性分析的cnn-lstm突发故障预警方法
CN112632845B (zh) 基于数据的小型反应堆在线故障诊断方法、介质及设备
CN109737045A (zh) 应用于芯片生产的空压机故障预警方法、系统及相关装置
Li et al. Application of Big Data Technology in Evaluation of Operating State of High-Pressure Hydrogenation Heat Exchanger
CN117370818B (zh) 基于人工智能的供排水管网智能诊断方法及智慧环保系统
CN106933097B (zh) 一种基于多层优化pcc-sdg的化工过程故障诊断方法
CN107220712B (zh) 一种基于马尔科夫过程的换流阀状态监测和检修决策的方法及系统
CN117744859A (zh) 基于自适应双控制策略的风电机组发电机故障预警方法
WO2019178948A1 (zh) 流程工业过程的一种多层模式监控方法
Shojaee et al. Applying SVSSI sampling scheme to np-chart to decrease the time of detecting shifts using Markov chain approach and Monte Carlo simulation
Peng et al. ASCS online fault detection and isolation based on an improved MPCA
CN116483865A (zh) 一种具有优先级的液压故障诊断专家系统
Knopik et al. Statistical analysis of parameters of rail vehicles
Hu et al. A multi-feature-based fault diagnosis method based on the weighted timeliness broad learning system
CN112306035B (zh) 化工过程运行故障的诊断系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18910641

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18910641

Country of ref document: EP

Kind code of ref document: A1