WO2023184764A1 - 一种基于粗糙集和证据理论的故障诊断方法及系统 - Google Patents

一种基于粗糙集和证据理论的故障诊断方法及系统 Download PDF

Info

Publication number
WO2023184764A1
WO2023184764A1 PCT/CN2022/102924 CN2022102924W WO2023184764A1 WO 2023184764 A1 WO2023184764 A1 WO 2023184764A1 CN 2022102924 W CN2022102924 W CN 2022102924W WO 2023184764 A1 WO2023184764 A1 WO 2023184764A1
Authority
WO
WIPO (PCT)
Prior art keywords
attribute
decision
data
attributes
condition
Prior art date
Application number
PCT/CN2022/102924
Other languages
English (en)
French (fr)
Inventor
贾宝柱
Original Assignee
广东海洋大学
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 广东海洋大学 filed Critical 广东海洋大学
Publication of WO2023184764A1 publication Critical patent/WO2023184764A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION 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
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION 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
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance

Definitions

  • the invention belongs to the field of mechanical equipment fault diagnosis, and specifically relates to a fault diagnosis method and system based on rough sets and evidence theory.
  • Fault diagnosis technology with data as the core is one of the key technologies to realize intelligent operation and maintenance of ship power systems.
  • operation monitoring data Through operation monitoring data, the operating status of equipment and systems is predicted, judged and identified, and based on this, it is determined whether they are in the normal operating state range. .
  • Long-term monitoring data of ship power systems usually have typical industrial big data characteristics such as large capacity, high latitude, and low effective information density.
  • Conventional methods are difficult to effectively mine potential feature information in the data, and are difficult to be directly used to support managers' operation, maintenance, and repair decisions. , it is necessary to find an effective data processing method and evaluate the effectiveness of the data processing results to support intelligent operation and maintenance decisions.
  • Rough set theory can explore the potential information hidden in the data without relying on prior knowledge of the data, providing a more feasible mathematical mechanism for acquiring knowledge in massive data.
  • the reasoning ability of rough set theory is weak, and it is highly complementary to other theories that deal with uncertain or imprecise problems such as probability theory, fuzzy mathematics, and evidence theory.
  • Rough sets are combined with information fusion of evidence theory. The method has important research significance in the fields of fault diagnosis, multi-sensor data fusion, condition assessment, data mining and pattern recognition.
  • Evidence theory can provide strong evidence support for the characterization and fusion of uncertain information at the decision-making level, but there is a large influence of subjective factors when determining the basic probability. At the same time, in the process of evidence combination, it will cause focal points due to the high dimensionality of the inference rules. Explosion, generally before using evidence theory for decision-level data fusion, it is necessary to make necessary reductions on the original data to reduce the data dimension. The advantages of rough sets in data processing and the reasoning ability of evidence theory can form a good complementary advantage, obtain more objective reasoning results in massive data processing, and can reduce the subjectivity and processing difficulty in evidence combination to a certain extent.
  • the purpose of this invention is to provide a fault diagnosis method and system based on rough sets and evidence theory, making full use of the attribute reduction ability of rough sets in the process of processing high-dimensional data, and the D-S evidence theory in the decision-level evidence synthesis process.
  • the confidence calculation method is used to overcome the defect of excessive subjectivity in the basic confidence allocation of decision-making evidence.
  • the present invention provides the following solutions:
  • a fault diagnosis method based on rough sets and evidence theory including the following steps:
  • the data decision table is represented as S ;
  • S5. According to the reduction decision table, obtain the basic confidence level of all condition attributes in the reduction decision table for each decision attribute.
  • the basic confidence level is used for fault diagnosis.
  • the original data type in the original data table in S1 is a continuous condition attribute
  • the original data is discretized according to the standard parameter range; after the normal interval range data is discretized, the condition attribute code is set to 1, which is lower than If the lower limit of the normal interval range is set, the condition attribute code is set to 0, and if it is higher than the upper limit of the normal interval range, the condition attribute code is set to 2.
  • the discretization process adopts the parameter index range under rated working conditions.
  • the parameters are determined according to the corresponding equipment operating point when the data is collected and combined with the variable working condition parameter change curve. the normal range.
  • the decision attribute set in S2 is a known fault state, and the inference relationship between it and the condition attributes is a non-linear or non-one-to-one mapping relationship. For situations where there is a simple one-to-one mapping relationship between condition attributes and decision attributes, Decision-making attributes are eliminated directly.
  • the attribute reduction processing method in S4 includes: attribute reduction processing based on equivalent attribute definitions of rough sets, and eliminating redundant attributes and related data information in the data decision table.
  • the equivalent attributes are the principle of simplifying the redundant condition attributes and related information existing in the data decision table while keeping the classification ability of the knowledge base unchanged;
  • x is a sample data vector, and its data dimension depends on the input condition attribute dimension n of the original collected data;
  • the method for obtaining the basic confidence level in S5 includes: adopting a basic confidence level allocation method based on evidence confidence level.
  • the basic confidence level is calculated as follows:
  • the invention also provides a fault diagnosis system based on rough sets and evidence theory, including: condition attribute module, decision attribute module, decision table module, processing module, and calculation module;
  • the conditional attribute module is used to obtain a conditional attribute set based on the parameter name and data category of the labeled data in the original data table;
  • the decision-making attribute module is used to obtain a decision-making attribute set according to the common fault status of the system
  • the decision table module is used to establish a data decision table based on the input condition attributes and decision attributes of the i-th original collected data sample under different working conditions based on the condition attribute set and the decision attribute set;
  • the processing module is used to perform attribute reduction processing on the data decision table to obtain a reduction decision table
  • the calculation module is used to obtain the basic confidence of each decision attribute of all condition attributes in the reduction decision table according to the reduction decision table.
  • the proposed evidence confidence calculation method makes full use of all the information in the decision table, and uses the evidence confidence as a conditional basic confidence distribution function to be more consistent with the actual situation, thereby improving the objectivity and accuracy of the decision conclusion.
  • Figure 1 is a schematic flow chart of a fault diagnosis method based on rough sets and evidence theory according to the present invention.
  • a fault diagnosis method based on rough sets and evidence theory includes the following steps:
  • the data decision attribute table is expressed as S;
  • S5. According to the reduction decision table, obtain the basic confidence level of all condition attributes in the reduction decision table for each decision attribute.
  • the basic confidence level is used for fault diagnosis.
  • the original data type in the original data table described in S1 is a continuous condition attribute
  • the original data is discretized according to the standard parameter range; after the normal interval range data is discretized, the condition attribute code is set to 1, which is lower than the normal interval range. If the lower limit is, set the condition attribute code to 0, and if it is higher than the upper limit of the normal range, set the condition attribute code to 2.
  • the discretization processing standard adopts the parameter index range under rated working conditions.
  • the normality of the parameters is determined based on the corresponding equipment working point when the data is collected and combined with the variable working condition parameter change curve. Interval range.
  • the decision attribute set described in S2 is the fault state of the known equipment or system, and the inference relationship between it and the condition attributes is a nonlinear or non-one-to-one mapping relationship. For situations where there is a simple one-to-one mapping relationship between condition attributes and decision attributes, Decision-making attributes are eliminated directly.
  • the method of attribute reduction processing described in S4 includes: attribute reduction processing based on equivalent attribute definitions of rough sets, and eliminating redundant attributes and related data information in the data decision table.
  • the equivalent attributes are the principle of simplifying the redundant condition attributes and related information existing in the data decision table while keeping the classification ability of the knowledge base unchanged;
  • x is a sample data vector, and its data dimension depends on the input condition attribute dimension n of the original collected data;
  • the method of obtaining the basic confidence in S5 includes: adopting a basic confidence allocation method based on evidence confidence.
  • the calculation method of the basic confidence is as follows:
  • a fault diagnosis method for ship power systems based on integrated learning which is characterized by including the following steps:
  • conditional attribute set C ⁇ c 1 , c 2 , c 3 ,..., c n ⁇ , where n is the input condition attribute Dimensions;
  • the ship central cooling system condition attribute set C contains a total of 26 input condition attributes, which are: c 1 - seawater pump inlet pressure, c 2 - seawater pump outlet pressure, c 3 - central cooler seawater inlet temperature, c 4 - Central cooler seawater outlet temperature, c 5 - Central cooler fresh water inlet temperature, c 6 - Low temperature fresh water pump inlet pressure, c 7 - Low temperature fresh water pump outlet pressure, c 8 - Oil cooler oil inlet pressure, c 9 - Oil outlet pressure of oil cooler, c 10 - Air inlet and outlet pressure difference of air cooler, c 11 - Low temperature fresh water temperature, c 12 - Fresh water outlet temperature of oil cooler, c 13 Oil inlet temperature of oil cooler, c 14 - Oil cooler oil outlet temperature, c 15 - Air cooler cooling water outlet temperature, c 16 - Air cooler air inlet temperature, c 17 - Air cooler air outlet temperature, c 18 - Main engine cylinder jacket water cooling pump inlet Pressure, c 19
  • the original data table U ⁇ x 1 ,..., x 30 ⁇ is discretized to obtain a discrete data table of condition attributes.
  • the decision attribute set D ⁇ d 1 , d 2 , d 3 ,..., d m ⁇ according to the common fault status of the system, where m is the dimension of the output decision attribute; the decision attribute vector set is the known fault
  • the inference relationship between the state and the input condition attributes is usually a non-linear or non-one-to-one mapping relationship. Decision attributes that have a simple one-to-one mapping relationship between condition attributes and decision attributes can be directly eliminated;
  • equivalent attribute classes obtained through equivalent attribute reduction are as follows:
  • ⁇ 1 ⁇ c 1 , c 3 , c 6 , c 8 , c 10 , c 13 , c 16 , c 18 , c 20 , c 21 ⁇ ;
  • ⁇ 2 ⁇ c 2 , c 11 ⁇ ;
  • ⁇ 3 ⁇ c 4 , c 5 ⁇ ;
  • ⁇ 4 ⁇ c 15 , c 17 ⁇ ;
  • ⁇ 5 ⁇ c 22 , c 23 ⁇ ;
  • ⁇ 6 ⁇ c 23 , c 26 ⁇ ;
  • S46 Calculate each condition attribute from high to low in approximate quality in C" according to the S44-S45 loop, and obtain the final reduction condition attribute set R and the reduction decision table.
  • the reduction condition attribute set R ⁇ c 2 , c 7 , c 9 , c 25 ⁇ obtained after attribute reduction;
  • S5. According to the reduction decision table, obtain the basic confidence level of all condition attributes in the reduction decision table for each decision attribute.
  • the basic confidence level is used for fault diagnosis.
  • the present invention provides a basic confidence allocation method based on evidence confidence, which is used to solve the subjective defective problem of the traditional basic confidence allocation method that only considers attribute weights.
  • the basic calculation method is as follows:
  • the invention also provides a fault diagnosis system based on rough sets and evidence theory, including: condition attribute module, decision attribute module, decision table module, processing module, and calculation module;
  • the conditional attribute module is used to obtain a conditional attribute set based on the parameter name and data category of the labeled data in the original data table;
  • the decision-making attribute module is used to obtain a decision-making attribute set according to the common fault status of the system
  • the decision table module is used to establish a data decision table based on the input condition attributes and decision attributes of the i-th original collected data sample under different working conditions based on the condition attribute set and the decision attribute set;
  • the processing module is used to perform attribute reduction processing on the data decision table to obtain a reduction decision table
  • the calculation module is used to obtain the basic confidence of each decision attribute of all condition attributes in the reduction decision table according to the reduction decision table.

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • Quality & Reliability (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • General Business, Economics & Management (AREA)
  • Educational Administration (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Test And Diagnosis Of Digital Computers (AREA)

Abstract

一种基于粗糙集和证据理论的故障诊断方法及系统，其中的方法包括步骤：S1、根据原始数据表中带标签数据的参数名称和数据类别，得到条件属性集；S2、根据系统常见故障状态得到决策属性集；S3、基于条件属性集和决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表；S4、对数据决策表进行属性的约简处理，获得约简决策表；S5、根据约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度。由此，通过将粗糙集与D-S证据理论相结合，克服了决策证据在基本置信度分配的主观性过强的缺陷。

Description

一种基于粗糙集和证据理论的故障诊断方法及系统 技术领域
本发明属于机械装备故障诊断领域，具体涉及一种基于粗糙集和证据理论的故障诊断方法及系统。
背景技术
以数据为核心的故障诊断技术是实现船舶动力系统智能运维的关键技术之一，通过运行监测数据对设备及系统的运行状态进行预报、判断及识别，据此确定其是否处于正常运行状态区间。船舶动力系统长期监测数据通常具有容量大、纬度高、有效信息密度低等典型工业大数据特征，常规方法难以有效挖掘数据中潜在的特征信息，难以直接用于支持管理人员的运行维护及维修决策，需要寻找一种有效的数据处理手段并对数据处理结果进行有效性评估以支撑智能运维决策。
粗糙集理论可以在不借助对数据的先验知识的前提下发掘隐藏在数据中的潜在信息，为海量数据中知识获取提供了一种较为可行的数学机制。但在实际应用中，粗糙集理论推理能力较弱，与概率论、模糊数学和证据理论等其它处理不确定或不精确问题的理论有很强的互补性，以粗糙集结合证据理论的信息融合方法在故障诊断、多传感器数据融合、状态评估、数据挖掘及模式识别等领域具有重要研究意义。
证据理论可以在决策级为不确定信息的表征与融合提供有力的证据支持，但在基本概率确定时存在较大的主观因素影响，同时在证 据组合过程中会由于推理规则维度过高引起焦元爆炸，一般在使用证据理论进行决策级数据融合前需要对原始数据进行必要的约简，以降低数据维度。粗糙集在数据处理环节的优势与证据理论的推理能力可以形成良好的优势互补，在海量数据处理中获得较为客观的推理结果，并可在一定程度上降低证据组合中的主观性及处理难度。
基于粗糙集和证据理论的信息融合方法目前有较多的研究成果，但是大部分研究成果仅针对具体应用对象特性提出了相应的数据融合方法，缺少能够推广应用的普适性方法。
发明内容
本发明的目的在于给出一种基于粗糙集和证据理论的故障诊断方法及系统，充分利用粗糙集在处理高维度数据过程中的属性约简能力，以及D-S证据理论在决策级证据合成过程中的置信度计算方法，以克服决策证据在基本置信度分配的主观性过强缺陷。
为实现上述目的，本发明提供了如下方案：
一种基于粗糙集和证据理论的故障诊断方法，包括如下步骤：
S1、根据原始数据表中带标签数据的参数名称和数据类别，得到条件属性集，所述条件属性集表示为C＝{c 1，c 2，c 3，...，c n}，其中n为输入的条件属性的维度；
S2、根据系统常见故障状态得到决策属性集，所述决策属性集表示为D＝{d 1，d 2，d 3，...，d m}，其中m为输出的决策属性的维度；
S3、基于所述条件属性集和所述决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数 据决策表，所述数据决策表表示为S；
S4、对所述数据决策表进行属性的约简处理，获得约简决策表；
S5、根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度，所述基本置信度用于进行故障诊断。
优选的，S1中所述原始数据表中的原始数据类型为连续的条件属性时，按照标准参数范围对原始数据进行离散化处理；正常区间范围数据离散化后设置条件属性码为1，低于正常区间范围下限则设置条件属性码为0，高于正常区间范围上限则设置其条件属性码为2。
优选的，所述离散化处理采用额定工况下参数指标范围，对于非额定工况下采集的参数数据，根据数据采集时对应的设备工况点，并结合变工况参数变化曲线来确定参数的正常区间范围。
优选的，S2中所述决策属性集为已知的故障状态，其与条件属性之间的推理关系为非线性或非一一映射关系，对于条件属性与决策属性间存在简单一一映射关系的决策属性直接剔除。
优选的，S4中所述属性的约简处理的方法包括：基于粗糙集的等价属性定义进行的属性约简处理，剔除数据决策表中的冗余属性及相关数据信息。
优选的，所述等价属性为在保持知识库分类能力不变的前提下，对数据决策表中存在的冗余条件属性及其相关信息进行简约的原则；
所述等价属性定义为：
Figure PCTCN2022102924-appb-000001
其中，x为样本数据向量，其数据维度取决于原始采集数据的输入条件属性维度n；X为所有样本数据向量构成的样本数据集， X＝{x 1，x 2，x 3，...，x k}，其中k为样本数据的数量；c i为条件属性，c i(x)为样本数据x在属性下c i的值，C＝{c 1，c 2，c 3，...，c n}为所有条件属性集合。
优选的，S5中获取所述基本置信度的方法包括：采用基于证据置信度的基本置信度分配方法。
优选的，所述基本置信度的计算方法如下：
Figure PCTCN2022102924-appb-000002
其中i＝1,2,...,n，Y i＝{y|y∈U^y D＝d i}，d i∈V D，U为原始数据表。
本发明还提供了一种基于粗糙集和证据理论的故障诊断系统，包括：条件属性模块、决策属性模块、决策表模块、处理模块、计算模块；
所述条件属性模块用于根据原始数据表中带标签数据的参数名称和数据类别，得到条件属性集；
所述决策属性模块用于根据系统常见故障状态得到决策属性集；
所述决策表模块用于基于所述条件属性集和所述决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表；
所述处理模块用于对所述数据决策表进行属性的约简处理，获得约简决策表；
所述计算模块用于根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度。
本发明的有益效果为：
(1)通过粗糙集方法与D-S证据理论相结合，充分利用粗糙集在处理高维度数据过程中的属性约简能力，以及D-S证据理论在决策级证据合成过程中的置信度计算方法，克服了决策证据在基本置信度分配的主观性过强缺陷；
(2)采用等价属性对输入条件属性进行约简，确定能够满足构建等价决策表所需的最小条件属性的数量，有利于减少决策级数据推理规则数量，避免焦元爆炸及其带来的计算负荷增加问题；
(3)提出的证据置信度计算方法，充分利用了决策表中的所有信息，将证据置信度作为条件基本置信度分配函数更加符合实际情况，从而提高了决策结论的客观性和准确性。
附图说明
为了更清楚地说明本发明的技术方案，下面对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本发明一种基于粗糙集和证据理论的故障诊断方法的流程示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中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，本领域普通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。
为使本发明的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本发明作进一步详细的说明。
实施例一
在本实施例一中，如图1所示，一种基于粗糙集和证据理论的故障诊断方法，包括如下步骤：
S1、根据原始数据表U中带标签数据的参数名称和数据类别，得到条件属性集，所述条件属性集表示为C＝{c 1，c 2，c 3，...，c n}，其中n为输入的条件属性的维度；
S2、根据系统常见故障状态得到决策属性集，所述决策属性集表示为D＝{d 1，d 2，d 3，...，d m}，其中m为输出的决策属性的维度；
S3、基于所述条件属性集和所述决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表，所述数据决策属性表表示为S；
S4、对所述数据决策表S进行属性的约简处理，获得约简决策表；
所述属性的约简处理方法如下：
S41、在决策表S＝(U,CUD,V,f)中，定义约简条件属性集
Figure PCTCN2022102924-appb-000003
S42、计算等价属性，保留每组等价属性中的任意一个属性，剔除冗余属性，得到约简后的等价属性集C’＝{Γ 1，Γ 2，...，Γ l}，其中l 为等价属性集中条件属性的个数，Γ i(i＝1，2，...，l)为等价属性类；
S43、计算C’中每一个每一个条件属性的近似质量，按从大到小排序为C”；
Figure PCTCN2022102924-appb-000004
为条件属性的近似质量；
Figure PCTCN2022102924-appb-000005
为决策表S中决策属性集D的C正域；
S44、选择C”中近似质量最大的属性ci，计算约简条件属性集R在加入条件属性c i后的近似质量
Figure PCTCN2022102924-appb-000006
如果得到的
Figure PCTCN2022102924-appb-000007
则R＝R∪{c i}，将c i从C”中剔除；
S45、如果
Figure PCTCN2022102924-appb-000008
则约简结束，否则返回S44；
S46、按照S44-S45循环计算C”中近似质量由高到低的每一个条件属性，得到最终约简条件属性集R及约简决策表；
S5、根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度，所述基本置信度用于进行故障诊断。
S1中所述原始数据表中的原始数据类型为连续的条件属性时，按照标准参数范围对原始数据进行离散化处理；正常区间范围数据离散化后设置条件属性码为1，低于正常区间范围下限则设置条件属性码为0，高于正常区间范围上限则设置其条件属性码为2。
所述离散化处理标准采用额定工况下参数指标范围，对于非额定工况下采集的参数数据，根据数据采集时对应的设备工况点，并结合变工况参数变化曲线来确定参数的正常区间范围。
S2中所述决策属性集为已知设备或系统的故障状态，其与条件属性之间的推理关系为非线性或非一一映射关系，对于条件属性与决 策属性间存在简单一一映射关系的决策属性直接剔除。
S4中所述属性的约简处理的方法包括：基于粗糙集的等价属性定义进行的属性约简处理，剔除数据决策表中的冗余属性及相关数据信息。
所述等价属性为在保持知识库分类能力不变的前提下，对数据决策表中存在的冗余条件属性及其相关信息进行简约的原则；
所述等价属性定义为：
Figure PCTCN2022102924-appb-000009
其中，x为样本数据向量，其数据维度取决于原始采集数据的输入条件属性维度n；X为所有样本数据向量构成的样本数据集，X＝{x 1，x 2，x 3，...，x k}，其中k为样本数据的数量；c i为条件属性，c i(x)为样本数据x在属性下c i的值，C＝{c 1，c 2，c 3，...，c n}为所有条件属性集合。
S5中获取所述基本置信度的方法包括：采用基于证据置信度的基本置信度分配方法，所述基本置信度的计算方法如下：
Figure PCTCN2022102924-appb-000010
其中i＝1,2,...,n，Y i＝{y|y∈U^y D＝d i}，d i∈V D。
实施例二
一种基于集成学习的船舶动力系统故障诊断方法，其特征在于，包括以下步骤：
S1、根据原始数据表U中带标签数据的参数名称和数据类别，得到条件属性集C＝{c 1，c 2，c 3，...，c n}，其中n为输入的条件属性的维度；
具体的，船舶中央冷却系统条件属性集C共包含26个输入条件属性，分别为：c 1-海水泵进口压力、c 2-海水泵出口压力、c 3-中央冷却器海水进口温度、c 4-中央冷却器海水出口温度、c 5-中央冷却器淡水进口温度、c 6-低温淡水泵进口压力、c 7-低温淡水泵出口压力、c 8-滑油冷却器滑油进口压力、c 9-滑油冷却器滑油出口压力、c 10-空冷器空气进出口压差、c 11-低温淡水温度、c 12-滑油冷却器淡水出口温度、c 13滑油冷却器滑油进口温度、c 14-滑油冷却器滑油出口温度、c 15-空冷器冷却水出口温度、c 16-空冷器空气进口温度、c 17-空冷器空气出口温度、c 18-主机缸套水冷却泵进口压力、c 19-主机缸套水冷却泵出口压力、c 20-缸套水冷却器缸套水进口压力、c 21-缸套水冷却器缸套水出口压力、c 22-缸套水冷却器低温淡水进口压力、c 23-缸套水冷却器低温淡水出口压力、c 24-主机缸套水出口温度、c 25-缸套水冷却器缸套水出口温度、c 26-缸套水冷却器冷却水出口温度。
根据系统说明书确定的各参数在额定工况下的离散化标准，对原始数据表U＝{x 1,...,x 30}进行离散化处理后得到条件属性的离散数据表。
S2、根据系统常见故障状态定义决策属性集D＝{d 1，d 2，d 3，...，d m}，其中m为输出的决策属性的维度；决策属性向量集为已知的故障状态，其与输入条件属性之间的推理关系通常为非线性或非一一映射关系。对于条件属性与决策属性间存在简单一一映射关系的决策属性可以直接剔除；
具体的，船舶中央冷却水系统决策属性集包括d 1＝1为正常状态、 d 2＝2为中央冷却器海水管路脏堵隐患状态、d 3＝3为低温淡水泵机械故障隐患状态、d 4＝4为滑油冷却器油路脏堵隐患状态、d 5＝5为高温淡水泵机械故障隐患状态。
S3、基于条件属性集和决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表S；
具体的，根据已知设备或系统的故障状态，确定包含输入条件属性集C中各条件属性C i(i＝1,2，...，26)与决策属性集D中各决策属性d j(i＝1,2，...，5)对应关系决策信息表。
S4、对数据决策表S进行属性的约简处理，获得约简决策表；
S41、在决策表S＝(U,CUD,V,f)中，定义约简条件属性集
Figure PCTCN2022102924-appb-000011
S42、计算等价属性，保留每组等价属性中的任意一个属性，剔除冗余属性，得到约简后的等价属性集C’＝{Γ 1，Γ 2，...，Γ l}，其中l为等价属性集中条件属性的个数，Γ i(i＝1，2，...，l)为等价属性类；
具体的，通过等价属性约简后获得的等价属性类如下：
Γ 1＝{c 1，c 3，c 6，c 8，c 10，c 13，c 16，c 18，c 20，c 21}；
Γ 2＝{c 2，c 11}；
Γ 3＝{c 4，c 5}；
Γ 4＝{c 15，c 17}；
Γ 5＝{c 22，c 23}；
Γ 6＝{c 23，c 26}；
约简后条件属性集C’：
C’＝{c 1，c 2，c 4，c 7，c 9，c 12，c 14，c 15，c 19，c 22，c 24，c 25}
S43、计算C’中每一个每一个条件属性的近似质量，按从大到小排序为C”；
Figure PCTCN2022102924-appb-000012
为条件属性的近似质量；
Figure PCTCN2022102924-appb-000013
为决策表S中决策属性集D的C正域；
具体的：
Figure PCTCN2022102924-appb-000014
Figure PCTCN2022102924-appb-000015
Figure PCTCN2022102924-appb-000016
S44、选择C”中近似质量最大的条件属性c i，计算初始条件属性集R在加入条件属性c i后的近似质量
Figure PCTCN2022102924-appb-000017
如果得到的
Figure PCTCN2022102924-appb-000018
则R＝R∪{c i}，将c i从C”中剔除；
S45、如果
Figure PCTCN2022102924-appb-000019
则约简结束，否则返回S44；
S46：按照S44-S45循环计算C”中近似质量由高到低的每一个条件属性，得到最终约简条件属性集R及约简决策表。
具体的，经过属性约简之后得到的约简条件属性集R＝{c 2，c 7，c 9，c 25}；
S5、根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度，所述基本置信度用于进行故障诊断。
本发明给出了基于证据置信度的基本置信度分配方法，用于解决传统仅考虑属性权重的基本置信度分配方法的主观缺陷性问题。其基本计算方法如下：
Figure PCTCN2022102924-appb-000020
其中i＝1，2，...，n，Y i＝{y|y∈U^y D＝d i}，d i∈V D。
具体的，根据约简决策表，应用S5给出的计算方法计算基本置信度，以获得约简决策表中所有条件属性对每个决策属性的基本置信度为：
m 1(d 1/c 2＝0)＝0，m 1(d 1/c 2＝1)＝3/11，m 1(d 1/c 2＝2)＝0；
m 1(d 2/c 2＝0)＝0，m 1(d 2/c 2＝1)＝1/22，m 1(d 2/c 2＝2)＝1；
m 1(d 3/c 2＝0)＝1/3，m 1(d 3/c 2＝1)＝5/22，m 1(d 3/c 2＝2)＝0；
m 1(d 4/c 2＝0)＝1/3，m 1(d 4/c 2＝1)＝5/22，m 1(d 4/c 2＝2)＝0；
m 1(d 5/c 2＝0)＝1/3，m 1(d 5/c 2＝1)＝5/22，m 1(d 5/c 2＝0)＝0；
m 2(d 1/c 7＝0)＝0，m 2(d 1/c 7＝1)＝3/11，m 2(d 1/c 7＝2)＝0；
m 2(d 2/c 7＝0)＝1/6，m 2(d 2/c 7＝1)＝5/22，m 2(d 2/c 7＝2)＝0；
m 2(d 3/c 7＝0)＝5/6，m 2(d 3/c 7＝1)＝1/22，m 2(d 3/c 7＝2)＝0；
m 2(d 4/c 7＝0)＝0，m 2(d 4/c 7＝1)＝2/11，m 2(d 4/c 7＝2)＝1；
m 2(d 5/c 7＝0)＝0，m 2(d 5/c 7＝1)＝3/11，m 2(d 5/c 7＝2)＝0；
m 3(d 1/c 9＝0)＝0，m 3(d 1/c 9＝1)＝3/11，m 3(d 1/c 9＝2)＝0；
m 3(d 2/c 9＝0)＝0，m 3(d 2/c 9＝1)＝3/11，m 3(d 2/c 9＝2)＝0；
m 3(d 3/c 9＝0)＝1/6，m 3(d 3/c 9＝1)＝5/22，m 3(d 3/c 9＝2)＝0；
m 3(d 4/c 9＝0)＝5/6，m 3(d 4/c 9＝1)＝1/22，m 3(d 4/c 9＝2)＝0；
m 3(d 5/c 9＝0)＝5/6，m 3(d 4/c 9＝1)＝1/22，m 3(d 4/c 9＝2)＝0；
m 3(d 5/c 9＝0)＝0，m 3(d 5/c 9＝1)＝2/11，m 3(d 5/c 9＝2)＝1；
m 4(d 1/c 25＝0)＝0，m 4(d 1/c 25＝1)＝6/19，m 4(d 1/c 25＝2)＝0；
m 4(d 2/c 25＝0)＝0，m 4(d 2/c 25＝1)＝2/19，m 4(d 2/c 25＝2)＝1；
m 4(d 3/c 25＝0)＝2/7，m 4(d 3/c 25＝1)＝4/19，m 4(d 3/c 25＝2)＝0；
m 4(d 4/c 25＝0)＝0，m 4(d 4/c 25＝1)＝6/19，m 4(d 4/c 25＝2)＝0；
m 4(d 5/c 25＝0)＝5/7，m 4(cd 5/c 25＝0)＝1/29，m 4(d 5/c 25＝2)＝0；
对基本置信度的物理意义可做如下理解：
以m 1(d 1/c 2)为例，当c 2＝0，即条件属性c 2离散后条件属性码为0时(对应海水泵出口压力低于正常范围低限值)，判定为决策属性为d 1(对应正常状态)的基本置信度为0；当c 2＝1，即条件属性c 2离散后条件属性码为1时(对应海水泵出口压力处于正常范围)，判定为决策属性d 1(对应正常状态)的基本置信度为3/11；当c 2＝2，即条件属性c 2离散后条件属性码为2时(对应海水泵出口压力高于正常范围高限值)，判定为决策属性为d 1(对应正常状态)的基本置信度为0。
实施例三
本发明还提供了一种基于粗糙集和证据理论的故障诊断系统，包括：条件属性模块、决策属性模块、决策表模块、处理模块、计算模块；
所述条件属性模块用于根据原始数据表中带标签数据的参数名称和数据类别，得到条件属性集；
所述决策属性模块用于根据系统常见故障状态得到决策属性集；
所述决策表模块用于基于所述条件属性集和所述决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表；
所述处理模块用于对所述数据决策表进行属性的约简处理，获得约简决策表；
所述计算模块用于根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度。
以上所述的实施例仅是对本发明优选方式进行的描述，并非对本发明的范围进行限定，在不脱离本发明设计精神的前提下，本领域普通技术人员对本发明的技术方案做出的各种变形和改进，均应落入本发明权利要求书确定的保护范围内。

Claims (9)

  1. 一种基于粗糙集和证据理论的故障诊断方法，其特征在于，包括如下步骤：
    S1、根据原始数据表中带标签数据的参数名称和数据类别，得到条件属性集，所述条件属性集表示为C＝{c 1，c 2，c 3，...，c n}，其中n为输入的条件属性的维度；
    S2、根据系统常见故障状态得到决策属性集，所述决策属性集表示为D＝{d 1，d 2，d 3，...，d m}，其中m为输出的决策属性的维度；
    S3、基于所述条件属性集和所述决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表，所述数据决策表表示为S；
    S4、对所述数据决策表进行属性的约简处理，获得约简决策表；
    S5、根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度，所述基本置信度用于进行故障诊断。
  2. 根据权利要求1所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，S1中所述原始数据表中的原始数据类型为连续的条件属性时，按照标准参数范围对原始数据进行离散化处理；正常区间范围数据离散化后设置条件属性码为1，低于正常区间范围下限则设置条件属性码为0，高于正常区间范围上限则设置其条件属性码为2。
  3. 根据权利要求2所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，所述离散化处理采用额定工况下参数指标范围，对于非额定工况下采集的参数数据，根据数据采集时对应的设备工况点， 并结合变工况参数变化曲线来确定参数的正常区间范围。
  4. 根据权利要求1所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，S2中所述决策属性集为已知的故障状态，其与条件属性之间的推理关系为非线性或非一一映射关系，对于条件属性与决策属性间存在简单一一映射关系的决策属性直接剔除。
  5. 根据权利要求1所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，S4中所述属性的约简处理的方法包括：基于粗糙集的等价属性定义进行的属性约简处理，剔除数据决策表中的冗余属性及相关数据信息。
  6. 根据权利要求5所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，所述等价属性为在保持知识库分类能力不变的前提下，对数据决策表中存在的冗余条件属性及其相关信息进行简约的原则；
    所述等价属性定义为：
    Figure PCTCN2022102924-appb-100001
    其中，x为样本数据向量，其数据维度取决于原始采集数据的输入条件属性维度n；X为所有样本数据向量构成的样本数据集，X＝{x 1，x 2，x 3，...，x k}，其中k为样本数据的数量；c i为条件属性，c i(x)为样本数据x在属性下c i的值，C＝{c 1，c 2，c 3，...，c n}为所有条件属性集合。
  7. 根据权利要求1所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，S5中获取所述基本置信度的方法包括：采用基于证据置信度的基本置信度分配方法。
  8. 根据权利要求1所述的基于粗糙集和证据理论的故障诊断方法，其特征在于，所述基本置信度的计算方法如下：
    Figure PCTCN2022102924-appb-100002
    其中，i＝1,2,...,n，Y i＝{y|y∈U∧y D＝d i}，d i∈V D，U为原始数据表。
  9. 一种基于粗糙集和证据理论的故障诊断系统，其特征在于，包括：条件属性模块、决策属性模块、决策表模块、处理模块、计算模块；
    所述条件属性模块用于根据原始数据表中带标签数据的参数名称和数据类别，得到条件属性集；
    所述决策属性模块用于根据系统常见故障状态得到决策属性集；
    所述决策表模块用于基于所述条件属性集和所述决策属性集，建立基于不同工况下第i条原始采集数据样本的输入条件属性与决策属性相对应的数据决策表；
    所述处理模块用于对所述数据决策表进行属性的约简处理，获得约简决策表；
    所述计算模块用于根据所述约简决策表，获得约简决策表中所有条件属性对每个决策属性的基本置信度。
PCT/CN2022/102924 2022-05-31 2022-06-30 一种基于粗糙集和证据理论的故障诊断方法及系统 WO2023184764A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210607154.1 2022-05-31
CN202210607154.1A CN115018307A (zh) 2022-05-31 2022-05-31 一种基于粗糙集和证据理论的故障诊断方法及系统

Publications (1)

Publication Number Publication Date
WO2023184764A1 true WO2023184764A1 (zh) 2023-10-05

Family

ID=83070741

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/102924 WO2023184764A1 (zh) 2022-05-31 2022-06-30 一种基于粗糙集和证据理论的故障诊断方法及系统

Country Status (2)

Country Link
CN (1) CN115018307A (zh)
WO (1) WO2023184764A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117909869A (zh) * 2024-03-20 2024-04-19 深圳大学 一种船舶事故因素识别方法、系统、终端及存储介质

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101216998A (zh) * 2008-01-11 2008-07-09 浙江工业大学 基于模糊粗糙集的证据理论城市交通流信息融合方法
CN102736562A (zh) * 2012-07-10 2012-10-17 北京信息科技大学 面向数控机床故障诊断与故障预报的知识库构建方法
CN108595635A (zh) * 2018-04-25 2018-09-28 中南大学 一种基于直觉模糊粗糙集的属性约简方法
CN108683663A (zh) * 2018-05-14 2018-10-19 中国科学院信息工程研究所 一种网络安全态势的评估方法及装置
CN111624985A (zh) * 2020-06-10 2020-09-04 上海工业自动化仪表研究院有限公司 燃气轮机控制系统传感器故障诊断方法
AU2020102424A4 (en) * 2020-09-25 2020-11-12 Beijing Institute Of Petrochemical Technology Hazardous chemical safety management assessment method
CN113689114A (zh) * 2021-08-23 2021-11-23 中国工商银行股份有限公司 一种信用度的确定方法、装置和设备
CN114491931A (zh) * 2021-12-17 2022-05-13 国网安徽省电力有限公司超高压分公司 数字孪生智能变电站系统故障的诊断方法及系统

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101216998A (zh) * 2008-01-11 2008-07-09 浙江工业大学 基于模糊粗糙集的证据理论城市交通流信息融合方法
CN102736562A (zh) * 2012-07-10 2012-10-17 北京信息科技大学 面向数控机床故障诊断与故障预报的知识库构建方法
CN108595635A (zh) * 2018-04-25 2018-09-28 中南大学 一种基于直觉模糊粗糙集的属性约简方法
CN108683663A (zh) * 2018-05-14 2018-10-19 中国科学院信息工程研究所 一种网络安全态势的评估方法及装置
CN111624985A (zh) * 2020-06-10 2020-09-04 上海工业自动化仪表研究院有限公司 燃气轮机控制系统传感器故障诊断方法
AU2020102424A4 (en) * 2020-09-25 2020-11-12 Beijing Institute Of Petrochemical Technology Hazardous chemical safety management assessment method
CN113689114A (zh) * 2021-08-23 2021-11-23 中国工商银行股份有限公司 一种信用度的确定方法、装置和设备
CN114491931A (zh) * 2021-12-17 2022-05-13 国网安徽省电力有限公司超高压分公司 数字孪生智能变电站系统故障的诊断方法及系统

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CHEN CHAO, CHEN XING-YUAN, YANG YING-JIE, WANG YONG-WEI: "System Security Assessment Based on Rough Set and D-S Evidence Theory", COMPUTER ENGINEERING, SHANGHAI JISUANJI XUEHUI, CN, vol. 39, no. 10, 15 October 2013 (2013-10-15), CN , pages 138 - 142, XP093097747, ISSN: 1000-3428, DOI: 10.3969/j.issn.1000-3428.2013.10.029 *
GUANG YANG, WU XIAOPING, SONG YEXIN: "Synthesized Fault Diagnosis Method Based on Fusion of Rough Sets and Evidence Theory", JOURNAL OF WUHAN UNIVERSITY OF TECHNOLOGY, vol. 31, no. 15, 15 August 2009 (2009-08-15), pages 105 - 110, XP093097750 *
YAO XIN-HUA, XU YUE-TONG, FU JIAN-ZHONG, CHEN ZI-CHEN: " Intelligent Fault Diagnosis of CNC Machine Tools based on Rough Set Theory", JOURNAL OF ZHEJIANG UNIVERSITY (ENGINEERING SCIENCE), GAIKAN BIANJIBU, HANGZHOU, CN, vol. 42, no. 10, 31 October 2008 (2008-10-31), CN , pages 1719 - 1724, XP009549443, ISSN: 1008-973X *

Also Published As

Publication number Publication date
CN115018307A (zh) 2022-09-06

Similar Documents

Publication Publication Date Title
Li et al. A novel evidential FMEA method by integrating fuzzy belief structure and grey relational projection method
US7941701B2 (en) Fuzzy classification approach to fault pattern matching
Zhu et al. A fuzzy rough number extended AHP and VIKOR for failure mode and effects analysis under uncertainty
Yang et al. The multiplicative consistency threshold of intuitionistic fuzzy preference relation
WO2023184764A1 (zh) 一种基于粗糙集和证据理论的故障诊断方法及系统
Wang et al. Data-driven risk assessment on urban pipeline network based on a cluster model
WO2022147853A1 (zh) 一种基于混合预测模型的复杂装备电源组故障预测方法
CN115187832A (zh) 一种基于深度学习与格拉姆角场图像的能源系统故障诊断方法
CN115795351B (zh) 一种基于残差网络和2d特征表示的电梯大数据风险预警方法
Yu et al. A novel FMEA approach for submarine pipeline risk analysis based on IVIFRN and ExpTODIM-PROMETHEE-II
Tian et al. Intelligent prediction and early warning of abnormal conditions for fluid catalytic cracking process
Gao et al. Mechanical equipment health management method based on improved intuitionistic fuzzy entropy and case reasoning technology
Sharma et al. Modeling system behavior for risk and reliability analysis using KBARM
Daher et al. New prognosis approach for preventive and predictive maintenance—Application to a distillation column
Chen et al. Interpretable mechanism mining enhanced deep learning for fault diagnosis of heating, ventilation and air conditioning systems
Dang et al. seq2graph: discovering dynamic dependencies from multivariate time series with multi-level attention
Amaitik et al. Developing a hierarchical fuzzy rule-based model with weighted linguistic rules: A case study of water pipes condition prediction
Shen et al. Decision-support system for infrastructure preservation
Fan Data mining model for predicting the quality level and classification of construction projects
CN114693175A (zh) 一种基于网源涉网试验的机组状态分析方法与系统
CN111832731B (zh) 一种多指标监测的油液不确定状态表征及故障诊断的方法
CN115936293A (zh) 一种基于pca的地铁施工安全事故风险评价方法
Liu et al. Decision support for maintenance management using Bayesian networks
Khanfri et al. New Hybrid MCDM Approach for an Optimal Selection of Maintenance Strategies: Results of a Case Study
CN108388232B (zh) 一种原油脱盐过程的运行模式故障监测方法

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 10202300001147

Country of ref document: CH

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

Ref document number: 22934587

Country of ref document: EP

Kind code of ref document: A1