技术领域technical field
本发明属于化工过程故障变量识别领域，特别涉及一种基于k近邻重构的工业过程故障变量识别方法。The invention belongs to the field of identification of chemical process fault variables, in particular to a method for identifying industrial process fault variables based on k-nearest neighbor reconstruction.
背景技术Background technique
对于过程监控和故障诊断问题，在传统的多元统计过程监控方法的框架下，主要采用贡献图和基于重构的贡献方法进行故障隔离。然而，这两种方法都存在故障干扰现象，即正常变量可能会受到故障变量的影响，使得正常变量对故障的贡献，从而导致误隔离。基于k近邻的变量贡献可以不受故障干扰的影响，但是它仅能给出故障变量大小排序，对于哪几个变量为故障变量的判别精度不高。另外，基于k近邻重构的故障变量识别方法对候选故障变量进行逐个估计并利用估计结果对故障样本完成重构，然后再对重构样本再检测实现故障识别，取得较好的识别效果和精度。但是，该方法识别速度非常慢，特别是对于变量维数较高的工业过程，无法及时给出变量识别结果，影响过程安全运行。本发明方法将解决这些问题。For process monitoring and fault diagnosis problems, under the framework of traditional multivariate statistical process monitoring methods, contribution graphs and reconstruction-based contribution methods are mainly used for fault isolation. However, both methods have fault interference phenomenon, that is, the normal variables may be affected by the fault variables, making the contribution of the normal variables to the fault, resulting in false isolation. The variable contribution based on k-nearest neighbors can not be affected by fault interference, but it can only give the order of fault variable size, and the discrimination accuracy of which variables are fault variables is not high. In addition, the fault variable identification method based on k-nearest neighbor reconstruction estimates the candidate fault variables one by one and uses the estimated results to complete the reconstruction of the fault samples, and then re-detects the reconstructed samples to realize fault identification, achieving better recognition effect and accuracy . However, the identification speed of this method is very slow, especially for industrial processes with high variable dimensions, the variable identification results cannot be given in time, which affects the safe operation of the process. The method of the present invention will solve these problems.
发明内容Contents of the invention
针对现有技术以上缺陷或改进需求，本发明提供一种基于k近邻重构的工业过程故障变量识别方法，在原始测量空间根据样本与近邻之间累计距离定义故障贡献指标，从而避免故障干扰问题，实现故障传感器准确地隔离。In view of the above defects or improvement needs of the prior art, the present invention provides a method for identifying fault variables in industrial processes based on k-nearest neighbor reconstruction, which defines the fault contribution index in the original measurement space according to the cumulative distance between the sample and the neighbors, thereby avoiding the problem of fault interference , to achieve accurate isolation of faulty sensors.
为实现上述目的，本发明通过以下技术方案实现：一种基于k近邻重构的工业过程故障变量识别方法，包括以下步骤：In order to achieve the above object, the present invention is achieved through the following technical solutions: a method for identifying industrial process fault variables based on k-nearest neighbor reconstruction, comprising the following steps:
1)利用多传感器数据采集系统收集工业过程正常工况运行下的监测数据，对其进行归一化构成正常数据矩阵其中，m表示监控变量的个数，n表示所总的采样数，表示第i个正常采样，表示实数域；1) Use the multi-sensor data acquisition system to collect the monitoring data under normal operating conditions of the industrial process, and normalize it to form a normal data matrix Among them, m represents the number of monitored variables, n represents the total number of samples, Denotes the i-th normal sample, represents the field of real numbers;
2)计算每个训练数据在正常数据矩阵X中的k近邻距离，并确定检测阈值：2) Calculate the k-nearest neighbor distance of each training data in the normal data matrix X, and determine the detection threshold:
2.1)从数据矩阵X中找每个采样xi的k个近邻2.1) Find the k nearest neighbors of each sample xi from the data matrix X
di,j＝||xi-xj||2,j＝1,…,n,j≠i (1)di,j =||xi -xj ||2 ,j=1,…,n,j≠i (1)
其中，||·||2表示l2范数，即欧式距离，k表示近邻个数，xj表示xi的第j个近邻，di,j表示xi和xj之间的欧式距离；Among them, ||·||2 represents the l2 norm, that is, the Euclidean distance, k represents the number of neighbors, xj represents the jth neighbor of xi , di,j represents the Euclidean distance betweenxi and xj ;
2.2)计算xi与其k个近邻xj,j＝1,…,k之间的平均距离：2.2) Calculate the average distance between xi and its k neighbors xj , j=1,...,k:
2.3)确定检测阈值对共n个数值按升序重新排列，并取第个作为检测阈值其中1-α表示置信水平。2.3) Determine the detection threshold right A total of n values are rearranged in ascending order, and the first as detection threshold where 1-α represents the confidence level.
3)对于工业过程在线采集测量数据首先利用X的均值和方差进行归一化处理得到然后根据以下步骤判断过程是否存在故障：3) For online collection of measurement data in industrial processes, first use the mean and variance of X to perform normalization processing to obtain Then judge whether there is a fault in the process according to the following steps:
3.1)根据公式(1)从正常数据矩阵X中找在线采样数据x的k个近邻；3.1) Find the k nearest neighbors of the online sampling data x from the normal data matrix X according to the formula (1);
3.2)根据公式(2)计算x与其k个近邻之间的平均距离；比较平均距离与检测阈值之间的大小，如果平均距离大于则说明工业过程发生故障；如果平均距离小于则说明工业过程正常运行。3.2) Calculate the average distance between x and its k neighbors according to formula (2); compare the average distance with the detection threshold The size between, if the average distance is greater than then the industrial process is faulty; if the average distance is less than It means that the industrial process is running normally.
4)将归一化在线采样数据x的k近邻累计距离的形式改写为4) The k-nearest neighbor cumulative distance of the normalized online sampling data x in the form of
其中，表示单位矩阵的第i行。in, represents the ith row of the identity matrix.
4.1)将x的k近邻累积距离分解m个分量之和，第i个分量为4.1) The cumulative distance of the k nearest neighbors of x Decompose the sum of m components, the i-th component is
4.2)将作为变量i对故障的贡献，对所有变量贡献由大到小进行排列：4.2) Will As the contribution of variable i to the fault, the contributions of all variables are arranged in descending order:
其中，vi∈{1,…,m}表示变量vi是第i个贡献最大的。Among them, vi ∈ {1,...,m} means that the variable vi is the i-th one with the largest contribution.
5)设定循环变量p＝1。5) Set loop variable p=1.
6)将候选故障变量vp加入候选变量集合v＝{vi,i＝1,…,p}中。根据集合v中的候选变量，去掉数据矩阵X以及故障数据x中对应集合v中候选变量，得到约简的正常数据矩阵和约简的故障数据6) Add the candidate fault variable vp into the candidate variable set v={vi , i=1, . . . , p}. According to the candidate variables in the set v, the data matrix X and the candidate variables in the corresponding set v in the fault data x are removed to obtain the reduced normal data matrix and reduced fault data
7)从约简的正常数据矩阵X(v)中找x(v)的k个近邻，根据这k个近邻估计集合v中候选故障变量的值：7) Find k neighbors of x(v) from the reduced normal data matrix X(v) , and estimate the value of the candidate fault variable in the set v according to these k neighbors:
其中，是归一化的权重，dl是x(v)与其第i个近邻之间的欧式距离。是矩阵X(v)中第Nl(x(v))列采样数据，表示的第vi个分量。in, is the normalized weight and dl is the Euclidean distance between x(v) and its i-th neighbor. is the sampling data of column Nl (x(v) ) in the matrix X(v) , express The vi -th component of .
8)利用估计值替换故障数据x对应分量，得到重构样本8) Utilize estimates Replace the corresponding component of the fault data x to obtain the reconstructed sample
9)根据公式(2)计算重构样本的平均k近邻距离如果则集合v中的变量被识别为故障变量；如果令p＝p+1并重复步骤2)至8)直至识别出所有故障变量。9) Calculate the reconstructed sample according to formula (2) The average k-nearest neighbor distance if Then the variables in the set v are identified as failure variables; if Let p=p+1 and repeat steps 2) to 8) until all fault variables are identified.
本发明所构思的以上技术方案与现有技术相比，有益效果是：本发明提出的工业过程故障变量识别方法不受正常变量对于分析故障变量的干扰影响，同时利用变量贡献的可靠排序加速故障变量识别过程，可以使得故障变量识别速度。另外，所提出方法适用于非线性、非高斯过程故障识别。从而保证了复杂工业过程可靠运行。Compared with the prior art, the above technical solution conceived by the present invention has the beneficial effect that the identification method of industrial process fault variables proposed by the present invention is not affected by the interference of normal variables on the analysis of fault variables, and at the same time, the reliable sorting of variable contributions is used to accelerate faults The variable identification process can make the faulty variable identification speed. In addition, the proposed method is suitable for fault identification in nonlinear and non-Gaussian processes. This ensures the reliable operation of complex industrial processes.
附图说明Description of drawings
图1.本发明方法的流程框图。Fig. 1. Flow diagram of the method of the present invention.
图2.实施例1中故障数据的k近邻变量贡献。Figure 2. k-Nearest Neighbor Variable Contributions for Failure Data in Example 1.
图3.实施例1中故障重构结果。Figure 3. The results of fault reconstruction in Example 1.
图4.实施例2中故障数据的k近邻变量贡献。Figure 4. k-Nearest Neighbor Variable Contributions for Fault Data in Example 2.
图5.实施例2中故障重构结果。Fig. 5. Results of fault reconstruction in Example 2.
图6.实施例3中故障数据的k近邻变量贡献。Figure 6. k-Nearest Neighbor Variable Contributions for Fault Data in Example 3.
图7.实施例3中单变量故障重构结果。Figure 7. Results of univariate fault reconstruction in Example 3.
图8.实施例3中多变量故障重构结果。Figure 8. Multivariate fault reconstruction results in Example 3.
具体实施方式Detailed ways
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本发明，并不用于限定本发明。In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.
本发明提出的一种基于k近邻重构的工业过程故障变量识别方法，其流程框图如图1所示，包括以下各步骤：A kind of industrial process fault variable identification method based on k-nearest neighbor reconstruction proposed by the present invention, its flow chart as shown in Figure 1, comprises the following steps:
1)采集工业过程正常工况运行下的数据，对其进行归一化构成正常数据矩阵其中，m为变量个数，n为样本个数；1) Collect data under normal operating conditions of the industrial process, and normalize it to form a normal data matrix Among them, m is the number of variables, n is the number of samples;
2)计算每个训练数据在正常数据矩阵X中的k近邻距离，并确定检测阈值2) Calculate the k-nearest neighbor distance of each training data in the normal data matrix X, and determine the detection threshold
3)对于工业过程在线运行采集数据，首先利用X的均值和方差进行归一化处理得到然后从正常数据矩阵X中找在线样本x的k个近邻；计算x与其k个近邻之间的平均距离；比较平均距离与检测阈值之间的大小，如果平均距离大于则说明工业过程发生故障；3) For the data collected during online operation of the industrial process, first use the mean and variance of X to perform normalization processing to obtain Then find the k neighbors of the online sample x from the normal data matrix X; calculate the average distance between x and its k neighbors; compare the average distance with the detection threshold The size between, if the average distance is greater than then the industrial process is malfunctioning;
4)计算样本x对故障的贡献，对所有变量贡献进行降序排列4) Calculate the contribution of sample x to failure, and sort all variable contributions in descending order
5)设定循环变量p＝1。5) Set loop variable p=1.
6)将候选故障变量vp加入候选变量集合v＝{vi,i＝1,…,p}中。根据集合v中的候选变量，去掉数据矩阵X以及故障数据x中对应集合v中候选变量，得到约简的正常数据矩阵和约简的故障数据6) Add the candidate fault variable vp into the candidate variable set v={vi , i=1, . . . , p}. According to the candidate variables in the set v, the data matrix X and the candidate variables in the corresponding set v in the fault data x are removed to obtain the reduced normal data matrix and reduced fault data
7)从约简的正常数据矩阵X(v)中找x(v)的k个近邻，根据这k个近邻估计集合v中候选故障变量的值；7) Find k neighbors of x(v) from the reduced normal data matrix X(v) , and estimate the value of the candidate fault variable in the set v according to these k neighbors;
8)计算重构样本的平均k近邻距离如果则集合v中的变量被识别为故障变量；如果令p＝p+1并重复步骤2)至8)直至识别出所有故障变量。8) Calculating reconstructed samples The average k-nearest neighbor distance if Then the variables in the set v are identified as failure variables; if Let p=p+1 and repeat steps 2) to 8) until all fault variables are identified.
实施例Example
采用本发明提供的工业过程k近邻故障变量识别方法，对数值仿真实例以及田纳西-伊斯曼连续化工过程进行故障识别；Adopt the industrial process k-nearest neighbor fault variable identification method provided by the present invention to carry out fault identification to numerical simulation examples and Tennessee-Eastman continuous chemical process;
实施例1中的，数值仿真实例具体描述为：In Embodiment 1, the numerical simulation example is specifically described as:
其中，ei～N(0,0.012),i＝1,…,7为白噪声，七个监控变量xi,i＝1,…,7是由两个潜变量s1和s2生成，它们分别服从均匀分布s1～U(-10,-7)和高斯分布s2～N(-15,1)；总共500个正常样本产生组成数据矩阵X；另外500个样本用于产生故障样本，故障为变量1和变量2分别从第301时刻施加幅值为4的恒偏差。Among them, ei ～N(0,0.012 ), i=1,...,7 are white noises, and the seven monitoring variables xi , i=1,...,7 are generated by two latent variables s1 and s2 , they obey the uniform distribution s1 ～U(-10,-7) and the Gaussian distribution s2 ～N(-15,1); a total of 500 normal samples are used to generate the data matrix X; the other 500 samples are used to generate faults In the sample, the fault is that variable 1 and variable 2 respectively impose a constant deviation with an amplitude of 4 from the 301st moment.
采用本实施例提供的故障变量识别方法，对上述数值仿真实例进行故障变量识别的具体过程如下：Using the fault variable identification method provided in this embodiment, the specific process of fault variable identification for the above numerical simulation example is as follows:
1)对正常样本进行归一化，构成正常数据矩阵1) Normalize the normal samples to form a normal data matrix
2)计算每个训练数据在正常数据矩阵X中的k近邻距离，此例中k＝15，并确定检测阈值2) Calculate the k-nearest neighbor distance of each training data in the normal data matrix X, in this example k=15, and determine the detection threshold
3)对故障数据，首先利用X的均值和方差进行归一化处理，然后从正常数据矩阵X中找故障数据的k个近邻；计算x与其k个近邻之间的平均距离；比较平均距离与检测阈值之间的大小，平均距离大于说明工业过程发生故障；3) For the fault data, first use the mean value and variance of X to perform normalization processing, and then find the k neighbors of the fault data from the normal data matrix X; calculate the average distance between x and its k neighbors; compare the average distance with detection threshold The size between, the average distance is greater than Explain that an industrial process has malfunctioned;
4)计算每个故障样本x对故障的贡献，对所有变量贡献进行降序排列4) Calculate the contribution of each failure sample x to the failure, and arrange all variable contributions in descending order
5)设定循环变量p＝1。5) Set loop variable p=1.
6)将候选故障变量vp加入候选变量集合v＝{vi,i＝1,…,p}中。根据集合v中的候选变量，去掉数据矩阵X以及故障数据x中对应集合v中候选变量，得到约简的正常数据矩阵和约简的故障数据6) Add the candidate fault variable vp into the candidate variable set v={vi , i=1, . . . , p}. According to the candidate variables in the set v, the data matrix X and the candidate variables in the corresponding set v in the fault data x are removed to obtain the reduced normal data matrix and reduced fault data
7)从约简的正常数据矩阵X(v)中找x(v)的k个近邻，根据这k个近邻估计集合v中候选故障变量的值；7) Find k neighbors of x(v) from the reduced normal data matrix X(v) , and estimate the value of the candidate fault variable in the set v according to these k neighbors;
计算重构样本的平均k近邻距离如果则集合v中的变量被识别为故障变量；如果令p＝p+1并重复步骤2)至8)直至识别出所有故障变量。Calculate reconstructed samples The average k-nearest neighbor distance if Then the variables in the set v are identified as failure variables; if Let p=p+1 and repeat steps 2) to 8) until all fault variables are identified.
图2是故障样本的变量贡献图，从图2中可以看到变量2和变量1的贡献明显高于其他变量。图3是重构样本再检测图，从图3的前7个子图中可以看出仅重构7个变量中的某一个是无法使重构样本变为正常，而图3中第8个子图(第三行、第二列)中呈现现有技术方案对变量2和变量1单独估计的结果也仍然有69％故障样本无法被重构恢复为正常，图3右下角最后一图为本发明方法所给出的结果，从子图中可以看出本发明方法可以将99.5％的故障样本通过重构变量2和变量1恢复正常，说明采用本方法提供的故障变量识别方法，可以准确的判定变量2和变量1为故障变量。Figure 2 is the variable contribution diagram of the failure sample. From Figure 2, it can be seen that the contribution of variable 2 and variable 1 is significantly higher than that of other variables. Figure 3 is a re-detection diagram of the reconstructed sample. From the first 7 subgraphs in Figure 3, it can be seen that only reconstructing one of the 7 variables cannot make the reconstructed sample normal, while the eighth subgraph in Figure 3 (Third row, second column) presents the results of separate estimation of variable 2 and variable 1 in the existing technical scheme, and 69% of the faulty samples cannot be reconstructed and restored to normal. The last figure in the lower right corner of Fig. 3 is the present invention From the results given by the method, it can be seen from the sub-graph that the method of the present invention can restore 99.5% of the fault samples to normal by reconstructing variable 2 and variable 1, which shows that the fault variable identification method provided by this method can accurately determine Variable 2 and variable 1 are fault variables.
实施例2采用与实施例1相同的500个正常样本，改变故障样本的故障类型为缓变故障，从第301采用时刻施加，幅值按照每时刻0.02的大小增长直至结束。对实施例2进行故障变量识别的具体过程与实施例1相同，图4和图5分别展示了变量贡献和故障变量识别结果。Embodiment 2 uses the same 500 normal samples as in Embodiment 1, and changes the fault type of the fault samples to slowly changing faults. It is applied from the 301st adoption time, and the amplitude increases by 0.02 per time until the end. The specific process of fault variable identification for embodiment 2 is the same as that of embodiment 1, and Fig. 4 and Fig. 5 show variable contribution and fault variable identification results respectively.
从图4中可以看出，变量2和变量1的贡献逐渐大于其他变量对故障的贡献，这与施加的故障情况一致，说明本发明提供的k近邻变量贡献能够提供可靠准确的变量贡献大小排序。从图5可以看出，现有技术仅重构某一个故障变量无法故障数据重构恢复正常，单独重构变量2和变量1仅能将75％的故障数据恢复正常。而本发明提供的方法可以将99.5％的故障数据恢复正常，可见本发明采用的故障变量识别方法，将变量2和变量1判定为故障变量，和实际情况相符。It can be seen from Figure 4 that the contribution of variable 2 and variable 1 is gradually greater than the contribution of other variables to the fault, which is consistent with the imposed fault situation, indicating that the k-nearest neighbor variable contribution provided by the present invention can provide a reliable and accurate ranking of variable contribution . It can be seen from FIG. 5 that in the prior art, only one fault variable can be reconstructed and the fault data cannot be restored to normal. Reconstructing variable 2 and variable 1 alone can only restore 75% of the fault data to normal. However, the method provided by the present invention can restore 99.5% of the fault data to normal. It can be seen that the fault variable identification method adopted by the present invention determines variable 2 and variable 1 as fault variables, which is consistent with the actual situation.
实施例3结合田纳西-伊斯曼化工过程数据来说明本发明方法的有效性。该标准测试实验平台由Downs和Vogel根据伊斯曼化学公司的一个实际化工联合反应过程所开发，其过程数据具有复杂非线性、强耦合和时变等特性，为监控方法提供了一个真实的工业过程。该过程由连续搅拌式反应釜、分凝器、汽液分离塔、离心式压缩机和汽提塔五个操作单元组成，包含了八种成分：A，B，C，D，E，F，G和H。其中，A，C，D，E四种气体进料成分和惰性气体成分B一起作为反应物，形成产物G和H，以及副产物F。田纳西-伊斯曼化工过程共有41个测量变量和12个控制变量，这里选取其中33个变量作为监控变量。正常训练样本数为960个，这里选取故障14为例说明本发明方法，与故障14直接相关的监控变量包括变量 9、变量21和变量32。Example 3 illustrates the effectiveness of the method of the present invention in combination with Tennessee-Eastman chemical process data. The standard test platform was developed by Downs and Vogel based on an actual chemical joint reaction process of Eastman Chemical Company. process. The process consists of five operating units: continuous stirring reactor, partial condenser, vapor-liquid separation tower, centrifugal compressor and stripping tower, and contains eight components: A, B, C, D, E, F, G and H. Among them, A, C, D, E four gas feed components and inert gas component B are used as reactants together to form products G and H, and by-product F. There are 41 measured variables and 12 controlled variables in the Tennessee-Eastman chemical process, 33 of which are selected as monitored variables here. The number of normal training samples is 960. Here, fault 14 is selected as an example to illustrate the method of the present invention. The monitoring variables directly related to fault 14 include variable 9, variable 21 and variable 32.
实施例3进行故障变量识别的具体过程与实施例1相同。图6展示了本发明方法中采用的k近邻变量贡献，从中可见变量9、变量21、变量32的贡献大小明显大于其他变量的贡献。图7和图8展示了故障重构结果，其中图7是仅对单变量重构的结果，从中可见仅重构单变量无法将故障数据恢复正常；图8对贡献最大的3个变量，即变量9、变量21和变量32进行重构，其中图8中上面的子图对这3个变量单独重构，结果显示仍有1.88％的故障数据未能被恢复正常，而本方面提供的方法对这3个变量同时估计并重构故障样本，结果显示仅有1.13％的故障数据未能被恢复正常，本发明提供的方法判定故障变量为变量9、变量21和变量32，这与实际情况相符，本发明仅用3次重构就可以完成故障变量识别，而现有类似技术在最佳情况仍需要529次(最差情况5456次)重构才能完成故障识别。说明了本发明方法具有较好的故障重构效果以及较快的故障变量识别速度。The specific process of identifying fault variables in Embodiment 3 is the same as Embodiment 1. Fig. 6 shows the contribution of the k-nearest neighbor variable used in the method of the present invention, from which it can be seen that the contribution of variable 9, variable 21, and variable 32 is significantly greater than that of other variables. Figure 7 and Figure 8 show the results of fault reconstruction, in which Figure 7 is the result of only single variable reconstruction, from which it can be seen that only reconstruction of single variable cannot restore the fault data to normal; Figure 8 has the largest contribution to the three variables, namely Variable 9, variable 21 and variable 32 are reconstructed, and the upper subgraph in Figure 8 reconstructs these 3 variables separately, the result shows that there are still 1.88% of fault data that cannot be restored to normal, and the method provided in this aspect Simultaneously estimate and reconstruct the fault sample for these 3 variables, the result shows that only 1.13% of the fault data cannot be restored to normal, and the method provided by the present invention determines that the fault variables are variable 9, variable 21 and variable 32, which is different from the actual situation Correspondingly, the present invention can complete fault variable identification with only 3 reconfigurations, while the existing similar technology still needs 529 reconfigurations (5456 times in worst case) to complete fault identification in the best case. It shows that the method of the present invention has better fault reconstruction effect and faster fault variable identification speed.
上述实施例用来解释说明本发明，而不是对本发明进行限制，在本发明的精神和权利要求的保护范围内，对本发明做出的任何修改和改变，都落入本发明的保护范围。The above-mentioned embodiments are used to illustrate the present invention, rather than to limit the present invention. Within the spirit of the present invention and the protection scope of the claims, any modification and change made to the present invention will fall into the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910476970.1ACN110308713A (en) | 2019-06-03 | 2019-06-03 | A Method for Identification of Industrial Process Fault Variables Based on k-Nearest Neighbor Reconstruction |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910476970.1ACN110308713A (en) | 2019-06-03 | 2019-06-03 | A Method for Identification of Industrial Process Fault Variables Based on k-Nearest Neighbor Reconstruction |
| Publication Number | Publication Date |
|---|---|
| CN110308713Atrue CN110308713A (en) | 2019-10-08 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910476970.1APendingCN110308713A (en) | 2019-06-03 | 2019-06-03 | A Method for Identification of Industrial Process Fault Variables Based on k-Nearest Neighbor Reconstruction |
| Country | Link |
|---|---|
| CN (1) | CN110308713A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112286168A (en)* | 2020-10-20 | 2021-01-29 | 湖州师范学院 | K nearest neighbor process monitoring method based on statistic pattern analysis |
| CN114371686A (en)* | 2021-12-29 | 2022-04-19 | 以萨技术股份有限公司 | Multi-working-condition process fault detection method and system based on local neighborhood standardization |
| CN114923715A (en)* | 2022-04-24 | 2022-08-19 | 扬州惠特科技有限公司 | Abnormal state detection method and system for homogenization and tackifying reaction kettle device |
| CN115544901A (en)* | 2022-11-28 | 2022-12-30 | 西南石油大学 | Intelligent thick oil working condition fault identification method for small sample |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104615123A (en)* | 2014-12-23 | 2015-05-13 | 浙江大学 | K-nearest neighbor based sensor fault isolation method |
| CN106649789A (en)* | 2016-12-28 | 2017-05-10 | 浙江大学 | Integrated semi-supervised Fisher's discrimination-based industrial process fault classifying method |
| CN106843195A (en)* | 2017-01-25 | 2017-06-13 | 浙江大学 | Based on the Fault Classification that the integrated semi-supervised Fei Sheer of self adaptation differentiates |
| CN109409425A (en)* | 2018-10-09 | 2019-03-01 | 宁波大学 | A kind of fault type recognition method based on neighbour's constituent analysis |
| CN109491338A (en)* | 2018-11-09 | 2019-03-19 | 南通大学 | A kind of relevant method for diagnosing faults of multimode procedure quality based on sparse GMM |
| CN109716303A (en)* | 2016-09-19 | 2019-05-03 | 应用材料公司 | Time series fault detection, fault classification and transition analysis using K-nearest neighbor and logistic regression methods |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104615123A (en)* | 2014-12-23 | 2015-05-13 | 浙江大学 | K-nearest neighbor based sensor fault isolation method |
| CN109716303A (en)* | 2016-09-19 | 2019-05-03 | 应用材料公司 | Time series fault detection, fault classification and transition analysis using K-nearest neighbor and logistic regression methods |
| US20200210873A1 (en)* | 2016-09-19 | 2020-07-02 | Applied Materials, Inc. | Time-series fault detection, fault classification, and transition analysis using a k-nearest-neighbor and logistic regression approach |
| CN106649789A (en)* | 2016-12-28 | 2017-05-10 | 浙江大学 | Integrated semi-supervised Fisher's discrimination-based industrial process fault classifying method |
| CN106843195A (en)* | 2017-01-25 | 2017-06-13 | 浙江大学 | Based on the Fault Classification that the integrated semi-supervised Fei Sheer of self adaptation differentiates |
| CN109409425A (en)* | 2018-10-09 | 2019-03-01 | 宁波大学 | A kind of fault type recognition method based on neighbour's constituent analysis |
| CN109491338A (en)* | 2018-11-09 | 2019-03-19 | 南通大学 | A kind of relevant method for diagnosing faults of multimode procedure quality based on sparse GMM |
| Title |
|---|
| GUOZHUWANG: "Fault diagnosis using kNN reconstruction on MRI variables", <JOURNAL OF CHEMOMETRICS>* |
| 周哲: "基于K邻近的复杂工业过程故障诊断方法研究", 《中国优秀博硕士学位论文全文数据库（博士） 信息科技辑》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112286168A (en)* | 2020-10-20 | 2021-01-29 | 湖州师范学院 | K nearest neighbor process monitoring method based on statistic pattern analysis |
| CN114371686A (en)* | 2021-12-29 | 2022-04-19 | 以萨技术股份有限公司 | Multi-working-condition process fault detection method and system based on local neighborhood standardization |
| CN114923715A (en)* | 2022-04-24 | 2022-08-19 | 扬州惠特科技有限公司 | Abnormal state detection method and system for homogenization and tackifying reaction kettle device |
| CN114923715B (en)* | 2022-04-24 | 2025-03-18 | 扬州惠特科技有限公司 | A method and system for detecting abnormal state of homogenization and viscosity increasing reactor device |
| CN115544901A (en)* | 2022-11-28 | 2022-12-30 | 西南石油大学 | Intelligent thick oil working condition fault identification method for small sample |
| CN115544901B (en)* | 2022-11-28 | 2023-02-21 | 西南石油大学 | A small-sample intelligent fault identification method for heavy oil operating conditions |
| Publication | Publication Date | Title |
|---|---|---|
| US11586913B2 (en) | Power equipment fault detecting and positioning method of artificial intelligence inference fusion | |
| EP3876056B1 (en) | Abnormality detection device and program | |
| CN110308713A (en) | A Method for Identification of Industrial Process Fault Variables Based on k-Nearest Neighbor Reconstruction | |
| CN111353482A (en) | LSTM-based fatigue factor recessive anomaly detection and fault diagnosis method | |
| CN118261584B (en) | Transformer state evaluation method and system based on multi-parameter data | |
| CN114563150B (en) | Bridge health online detection module generation method, detection method, tool box and device | |
| JP7238378B2 (en) | Abnormality detection device, abnormality detection program, and abnormality detection method | |
| CN104035431A (en) | Obtaining method and system for kernel function parameters applied to nonlinear process monitoring | |
| CN116957361B (en) | Ship task system health state detection method based on virtual-real combination | |
| CN112949735A (en) | Liquid hazardous chemical substance volatile concentration abnormity discovery method based on outlier data mining | |
| CN104615123B (en) | K-nearest neighbor based sensor fault isolation method | |
| CN120316691B (en) | Abnormal state early warning method and system for integrated pipeline corridor based on Internet of Things | |
| CN110956331A (en) | Method, system and device for predicting operation state of digital factory | |
| CN119805200B (en) | Intelligent diagnosis method and device for power distribution network switch faults and electronic equipment | |
| CN115659271A (en) | Sensor abnormality detection method, model training method, system, device, and medium | |
| CN118585756B (en) | Air quality monitoring method and system based on Internet of things | |
| CN118378178B (en) | Transformer fault identification method and system based on residual map convolution neural network | |
| CN119128407A (en) | A method and system for preprocessing arch dam deformation monitoring data | |
| CN110967184B (en) | Gearbox fault detection method and system based on vibration signal distribution characteristic recognition | |
| CN109840386B (en) | Damage identification method based on factor analysis | |
| US11783233B1 (en) | Detection and visualization of novel data instances for self-healing AI/ML model-based solution deployment | |
| CN117312972A (en) | Method for identifying health state of scraper conveyor speed reducer | |
| de Oliveira et al. | Noise-resilient feature selection for accelerometer-based guyed tower monitoring | |
| CN114936205A (en) | Feature screening method and device, storage medium and electronic equipment | |
| CN114660931A (en) | Method and system for diagnosing and identifying petrochemical process fault |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20191008 |