Polysilicon production equipment fault diagnosis method and device, server and storage medium

Abstract

The application provides a polysilicon production equipment fault diagnosis method and device, a server and a storage medium, and relates to the field of equipment fault diagnosis. The method comprises the following steps: acquiring target equipment collected sample data and first historical sample data within a preset time window; determining a first set composed of equipment parameters and a second set composed of equipment parameters and corresponding upstream and downstream process parameters; further determining the first robust distance of the equipment parameters according to the equipment mean vector and the equipment covariance matrix; determining the second robust distance of the equipment parameters and the corresponding upstream and downstream process parameters according to the overall mean vector and the overall covariance matrix; if the first robust distance is greater than the first robust distance threshold and / or the second robust distance is greater than the second robust distance threshold, the running state is determined to be an abnormal state. The comprehensiveness of the diagnosis and the accuracy of the diagnosis result are ensured, the stability of the benchmark during diagnosis is ensured, and the diagnosis efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of equipment fault diagnosis technology, and in particular to fault diagnosis methods, devices, servers and storage media for polysilicon production equipment. Background Technology

[0002] Equipment fault diagnosis and early warning are crucial in the large-scale chemical industry. This is not only to ensure the safe operation of equipment and timely maintenance under abnormal conditions, but also to guarantee the safe and stable operation of chemical processes. Integrating the condition monitoring of critical equipment with upstream and downstream processes facilitates a close link between equipment management and process control, thereby improving the equipment health management system and enhancing its reliability, usability, safety, and efficiency.

[0003] Generally, for various types of moving and static equipment, most patents only analyze equipment parameters (and mostly amplitude), without considering the influence of processes.

[0004] Currently, in polysilicon manufacturing processes, the condition monitoring methods for equipment such as compressors and heaters are based on threshold detection using expert experience (i.e., warning and interlock values ​​from the manufacturer), involving equipment parameters such as current and vibration intensity. However, these parameters are only individual parameters and do not integrate with process conditions. In reality, 82% of equipment failures are caused by problems with process conditions. Therefore, a holistic condition monitoring method is urgently needed for fault diagnosis and real-time monitoring.

[0005] To address the aforementioned issues, existing technologies have attempted to utilize the Earthworm Box system based on dynamic principal component dimensionality reduction. While this system facilitates intuitive observation of the stability of real-time data and the presence of anomalies, it cannot establish a stable baseline space based on historical data. Specifically, when data is updated in real time, the system can only update the overall score matrix, load matrix, and eigenvalues ​​by combining historical data and then taking the corresponding values ​​of the new samples. This affects computational speed and hinders the understanding of the principle by system maintenance personnel and staff in the control room and instrument workshop. Summary of the Invention

[0006] The technical objective of this application is to provide a method, apparatus, server, and storage medium for diagnosing faults in polysilicon production equipment, in order to solve the problem that current methods of fault diagnosis based on individual equipment parameters cannot perform rapid and accurate equipment fault diagnosis.

[0007] To address the aforementioned technical problems, this application provides a method for diagnosing faults in polysilicon production equipment, comprising:

[0008] The target device acquires the sample data to be analyzed collected at the target time point and the first historical sample data collected at multiple first historical time points respectively. The time difference between the first historical time point and the target time point is less than a preset time difference. The target time is the monitoring time point or any historical time point.

[0009] Based on the sample data to be analyzed and the first historical sample data, a first set consisting of preset equipment parameters and a second set consisting of the equipment parameters and their corresponding upstream and downstream process parameters are determined.

[0010] Empirical Mode Decomposition (EMD) is performed on the device parameters in the first set to obtain the third set;

[0011] Based on the third set, the equipment mean vector, and the equipment covariance matrix, a first robust distance of the equipment parameters is determined; and based on the second set, the overall mean vector, and the overall covariance matrix, a second robust distance of the equipment parameters and their corresponding upstream and downstream process parameters is determined, wherein the equipment mean vector and the equipment covariance matrix are pre-determined based on the equipment parameters in multiple second historical data sets, and the overall mean vector and the overall covariance matrix are pre-determined based on the equipment parameters and their corresponding upstream and downstream process parameters in multiple second historical data sets.

[0012] If the first robust distance is greater than a predetermined first robust distance threshold, and / or the second robust distance is greater than a predetermined second robust distance threshold, the operating state of the target device within the sampling time corresponding to the target time point is determined to be an abnormal state; otherwise, the operating state is determined to be a normal state.

[0013] Preferably, the method described above further includes:

[0014] Obtain multiple sets of the second historical sample data from the target device;

[0015] Based on the second historical sample data, a fourth set consisting of the equipment parameters and a fifth set consisting of the equipment parameters and their corresponding upstream and downstream process parameters are determined.

[0016] Empirical mode decomposition is performed on the device parameters in the fourth set to obtain the sixth set;

[0017] Based on the sixth set, the device mean vector, the device covariance matrix, and the third robust distance are determined; and based on the fifth set, the overall mean vector, the overall covariance matrix, and the fourth robust distance are determined.

[0018] Furthermore, the method described above also includes:

[0019] The first robust distance threshold is determined based on the third robust distance, and the second robust distance threshold is determined based on the fourth robust distance.

[0020] Specifically, as described above, after obtaining the original sample data, the method further includes:

[0021] According to a preset preprocessing strategy, the original sample data is preprocessed to obtain the preprocessed original sample data, wherein the original sample data is the sample data to be analyzed and the first historical sample data, or the second historical sample data;

[0022] The preprocessing includes at least one of the following:

[0023] Parameter validity analysis;

[0024] Principal component analysis;

[0025] Interpolation processing;

[0026] Correlation analysis;

[0027] Sample sparsity processing.

[0028] Specifically, in the method described above, after performing empirical mode decomposition on the device parameters, the method further includes:

[0029] The maximum number of references for the IMF components is determined based on the number of Intrinsic Mode Function (IMF) components obtained from the decomposition of the device parameters.

[0030] IMF components are supplemented based on the maximum reference quantity, wherein the value of the supplemented IMF component is 0.

[0031] Furthermore, the method described above also includes:

[0032] After determining the operating status, the sample data to be analyzed is displayed and stored according to the operating status and / or the target time point, and the operating status is output.

[0033] Specifically, in the case where the operating state is the abnormal state, the method described above further includes:

[0034] An early warning is issued based on the first robust distance and / or the second robust distance used to determine the abnormal state.

[0035] Another embodiment of this application provides a fault diagnosis device for polysilicon production equipment, including:

[0036] The first processing module is used to acquire the sample data to be analyzed collected by the target device at the target time point and the first historical sample data collected at multiple first historical time points respectively. The time difference between the first historical time point and the target time point is less than a preset time difference. The target time is the monitoring time point or any historical time point.

[0037] The second processing module is used to determine, based on the sample data to be analyzed and the first historical sample data, a first set consisting of preset equipment parameters and a second set consisting of the equipment parameters and their corresponding upstream and downstream process parameters.

[0038] The third processing module is used to perform empirical mode decomposition on the device parameters in the first set to obtain a third set.

[0039] The fourth processing module is configured to determine a first robust distance of the equipment parameters based on the third set, the equipment mean vector, and the equipment covariance matrix; and to determine a second robust distance of the equipment parameters and their corresponding upstream and downstream process parameters based on the second set, the overall mean vector, and the overall covariance matrix, wherein the equipment mean vector and the equipment covariance matrix are pre-determined based on the equipment parameters in multiple second historical data sets, and the overall mean vector and the overall covariance matrix are pre-determined based on the equipment parameters and their corresponding upstream and downstream process parameters in multiple second historical data sets.

[0040] The fifth processing module is used to determine that the operating state of the target device within the sampling time corresponding to the target time point is an abnormal state if the first robust distance is greater than a predetermined first robust distance threshold and / or the second robust distance is greater than a predetermined second robust distance threshold; otherwise, it determines that the operating state is a normal state.

[0041] Another embodiment of this application provides a server, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the polysilicon production equipment fault diagnosis method as described above.

[0042] Another embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the polysilicon production equipment fault diagnosis method as described above.

[0043] Compared with the prior art, the polysilicon production equipment fault diagnosis method, apparatus, server and storage medium provided in this application have at least the following beneficial effects:

[0044] By acquiring only the sample data corresponding to the required diagnostic time point and some historical data prior to that, the computational load is reduced; the diagnostic process involves equipment parameters and their upstream and downstream process parameters, ensuring the comprehensiveness of the diagnosis and the accuracy of the diagnostic results; each parameter set calculates a robust distance based on a unified mean vector and covariance matrix, and makes judgments based on a unified threshold, ensuring the stability of the baseline during diagnosis and improving diagnostic efficiency. Attached Figure Description

[0045] Figure 1 This is one of the flowcharts illustrating a fault diagnosis method for polysilicon production equipment.

[0046] Figure 2 The second flowchart illustrates the fault diagnosis method for polysilicon production equipment.

[0047] Figure 3 This is a timing diagram of the device parameters;

[0048] Figure 4 A schematic diagram showing the robust distances corresponding to the overall equipment parameters and their upstream and downstream process parameters;

[0049] Figure 5 The empirical mode decomposition results and robust distance diagram for the equipment parameters;

[0050] Figure 6 A schematic diagram showing the robust distances corresponding to the equipment parameters;

[0051] Figure 7 One of the structural schematic diagrams of a fault diagnosis device for polysilicon production equipment;

[0052] Figure 8 This is the second schematic diagram of a fault diagnosis method for polysilicon production equipment. Detailed Implementation

[0053] To make the technical problems, technical solutions, and advantages of this application clearer, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments. In the following description, specific details such as particular configurations and components are provided merely to aid in a comprehensive understanding of the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Furthermore, for clarity and brevity, descriptions of known functions and structures have been omitted.

[0054] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0055] In the various embodiments of this application, it should be understood that the sequence number of each process described below does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0056] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0057] In the embodiments provided in this application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.

[0058] See Figure 1 One embodiment of this application provides a method for diagnosing faults in polysilicon production equipment, including:

[0059] Step S101: Obtain the sample data to be analyzed collected by the target device at the target time point and the first historical sample data collected at multiple first historical time points respectively. The time difference between the first historical time point and the target time point is less than a preset time difference. The target time is the monitoring time point or any historical time point.

[0060] Step S102: Based on the sample data to be analyzed and the first historical sample data, determine a first set consisting of preset equipment parameters and a second set consisting of the equipment parameters and their corresponding upstream and downstream process parameters.

[0061] Step S103: Perform empirical mode decomposition on the device parameters in the first set to obtain a third set;

[0062] Step S104: Based on the third set, the equipment mean vector, and the equipment covariance matrix, determine the first robust distance of the equipment parameters; and based on the second set, the overall mean vector, and the overall covariance matrix, determine the second robust distance of the equipment parameters and their corresponding upstream and downstream process parameters, wherein the equipment mean vector and the equipment covariance matrix are pre-determined based on the equipment parameters in multiple second historical data sets, and the overall mean vector and the overall covariance matrix are pre-determined based on the equipment parameters and their corresponding upstream and downstream process parameters in multiple second historical data sets.

[0063] Step S105: If the first robust distance is greater than a predetermined first robust distance threshold, and / or the second robust distance is greater than a predetermined second robust distance threshold, determine that the operating state of the target device within the sampling time corresponding to the target time point is an abnormal state; otherwise, determine that the operating state is a normal state.

[0064] In this embodiment, when diagnosing faults in polysilicon production equipment, sampling points are set on the target equipment and the equipment corresponding to the upstream and downstream process parameters of the target equipment. Sampling is then performed according to the sampling period, allowing the acquisition of the corresponding data sample to be analyzed at the target time when fault diagnosis is needed, as well as the first historical sample data prior to that time. It should be noted that the solution of this application can diagnose faults on historical data corresponding to any historical event point, and can also diagnose the current data sample corresponding to the current monitoring time point. Furthermore, determining the first historical sample data based on the time point of the sample data to be analyzed ensures the accuracy of fault diagnosis. Simultaneously, the temporal relationship between the first historical sample data and the sample data to be analyzed helps avoid the large computational load caused by using all historical data samples each time. For example, if the first historical sample data and the sample data to be analyzed are located within a preset data time window, when the time point corresponding to the sample data to be analyzed shifts later, the first historical time point corresponding to the first historical sample data also shifts later.

[0065] After acquiring the sample data to be analyzed and the first historical sample data, the preset equipment parameters are separately formed into a first set, and the equipment parameters and their corresponding upstream and downstream process parameters are collectively formed into a second set. This allows for the analysis of both the equipment parameters and their corresponding upstream and downstream process parameters as a whole, avoiding the problem of incomplete diagnosis caused by analyzing only one aspect, which could lead to inaccurate diagnostic results. The equipment parameters are those that have a significant impact on the target equipment. Different target equipment has different corresponding equipment parameters. For example, in a specific embodiment, the target equipment is a quench tower circulating pump, and the equipment parameter is current.

[0066] After determining the first set, to reflect the stability characteristics of the target device itself, the device parameters need to be analyzed independently. Therefore, empirical mode decomposition (EMD) is performed on the device parameters in the first set, and a third set is obtained based on the decomposed IMF components for subsequent calculations. The EMD method involves obtaining a time-series sample of a parameter and gradually separating several IMF components, leaving a residual component. Since this EMD method is existing technology, it will not be elaborated upon here.

[0067] During the calculation, a first robust distance for the equipment parameters is obtained based on the pre-acquired equipment mean vector and equipment covariance matrix, as well as the aforementioned third set. Similarly, a second robust distance for the equipment parameters and their upstream and downstream process parameters is obtained based on the pre-acquired overall mean vector and overall covariance matrix, as well as the aforementioned second set. The robust distance is an improved determinant based on the fast covariance determinant algorithm, which is existing technology; therefore, the steps for obtaining the robust distance will not be elaborated further. It should be noted that the equipment mean vector, equipment covariance matrix, overall mean vector, and overall covariance matrix involved in this step are all pre-determined based on corresponding parameters from multiple second historical data sets. That is, in this embodiment, the same equipment mean vector, equipment covariance matrix, overall mean vector, and overall covariance matrix are used for calculation during diagnosis, which helps ensure the stability of the baseline during diagnosis, thereby improving computational efficiency.

[0068] After obtaining the first and second robust distances mentioned above, they are compared with their corresponding thresholds. If the first robust distance is greater than a predetermined first robust distance threshold, the target equipment itself is determined to be faulty. If the second robust distance is greater than a predetermined second robust distance threshold, the upstream and downstream process equipment of the target equipment is determined to be faulty. If both are greater than their corresponding thresholds, the target equipment and its upstream and downstream process equipment are determined to be faulty, thus enabling timely and comprehensive diagnosis of the target equipment. It should be noted that the first robust distance threshold in this step is related to the equipment mean vector and equipment covariance matrix obtained from the second historical data in the previous step, and the second robust distance threshold in this step is related to the overall mean vector and overall covariance matrix obtained from the second historical data in the previous step.

[0069] In summary, this embodiment reduces the computational load by acquiring only the sample data corresponding to the required diagnostic time point and some historical data prior to that point; the diagnostic process involves equipment parameters and their upstream and downstream process parameters, ensuring the comprehensiveness of the diagnosis and the accuracy of the diagnostic results; each parameter set calculates a robust distance based on a unified mean vector and covariance matrix, and makes judgments based on a unified threshold, ensuring the stability of the baseline during diagnosis and improving diagnostic efficiency.

[0070] To facilitate understanding by those skilled in the art, the following illustrates the steps for obtaining the device mean vector, device covariance matrix, and first robust distance threshold corresponding to the device parameters, as well as the overall mean vector, overall covariance matrix, and second robust distance threshold corresponding to the device parameters and their corresponding upstream and downstream process parameters, obtained in advance in the above embodiments.

[0071] See Figure 2 Preferably, the method described above further includes:

[0072] Step S201: Obtain multiple sets of the second historical sample data from the target device;

[0073] Step S202: Based on the second historical sample data, determine the fourth set consisting of the equipment parameters and the fifth set consisting of the equipment parameters and their corresponding upstream and downstream process parameters.

[0074] Step S203: Perform empirical mode decomposition on the device parameters in the fourth set to obtain the sixth set;

[0075] Step S204: Based on the sixth set, determine the device mean vector, the device covariance matrix, and the third robust distance; and based on the fifth set, determine the overall mean vector, the overall covariance matrix, and the fourth robust distance.

[0076] In one embodiment, it is preferable to acquire multiple second historical sample data of the target device. These multiple second historical sample data are preferably all historical sample data of the target device or other devices with the same operating conditions and equipment status as the target device. However, if there is too much historical sample data, it can be appropriately reduced based on factors such as data timeliness and environment.

[0077] After obtaining the second historical sample data, based on similar steps as described above, such as set partitioning and empirical mode decomposition, the required fifth and sixth sets are obtained, which will not be elaborated here.

[0078] After obtaining the fifth and sixth sets, based on the optimized Mahalanobis distance algorithm (i.e., the minimum covariance determinant algorithm), the device mean vector and device covariance matrix corresponding to the sixth set, and the overall mean vector and overall covariance matrix corresponding to the fifth set are obtained using the above sets. Specifically, the device mean vector is the vector formed by the time-series mean of all components of the device parameters after empirical mode decomposition; the overall mean vector is the vector formed by the time-series mean of the device parameters and their upstream and downstream process parameters; the device covariance matrix is ​​the covariance matrix formed by the overall covariance of all components of the device parameters after empirical mode decomposition; and the overall covariance matrix is ​​the covariance matrix formed by the overall covariance of the device parameters and their upstream and downstream process parameters. Since the minimum covariance determinant algorithm is existing technology, it will not be elaborated upon here.

[0079] After obtaining the equipment mean vector and equipment covariance matrix, the third robust distance corresponding to the current equipment parameters can be obtained by combining it with the sixth set. Similarly, after obtaining the overall mean vector and overall covariance matrix, the fourth robust distance corresponding to the current equipment parameters and their upstream and downstream process parameters can be obtained by combining it with the fifth set. Furthermore, by adding coefficients and / or detection ranges to the third and fourth robust distances, the first robust distance threshold and the second robust distance threshold can be obtained. In a specific embodiment, the determination value of the first robust distance threshold is: Where x is the third robust distance, p is related to the number of upstream and downstream process parameters of the equipment, and a is a preset coefficient, typically 0.975. The detection range includes 2σ, 3σ, and 6σ detection ranges, generally suitable for real-time monitoring based on healthy samples.

[0080] Specifically, as described above, after obtaining the original sample data, the method further includes:

[0081] According to a preset preprocessing strategy, the original sample data is preprocessed to obtain the preprocessed original sample data, wherein the original sample data is the sample data to be analyzed and the first historical sample data, or the second historical sample data;

[0082] The preprocessing includes at least one of the following:

[0083] Parameter validity analysis;

[0084] Principal component analysis;

[0085] Interpolation processing;

[0086] Correlation analysis;

[0087] Sample sparsity processing.

[0088] In this embodiment, to ensure that the data meets the computational requirements and avoids excessive computation, the acquired raw data is preprocessed based on user settings. It should be noted that the raw sample data in this embodiment refers to the sample data to be used in subsequent calculations. For example, in fault diagnosis, the raw sample data includes the sample data to be analyzed and the first historical sample data. When obtaining the device mean vector, device covariance matrix, overall mean vector, overall covariance matrix, first robust distance threshold, and second robust distance threshold, the raw sample data is the second historical data. Preprocessing methods include, but are not limited to: parameter validity analysis; principal component analysis; interpolation; correlation analysis; and sample sparsity processing. In actual use, one or more of these methods are selected for preprocessing according to user instructions.

[0089] The following is a brief description of the implementation methods of the various preprocessing methods described above.

[0090] Parameter validity analysis involves setting corresponding parameter thresholds (warning values ​​or cascading values) for the parameters collected at each sampling point. When the parameter value corresponding to the target parameter exceeds the corresponding parameter threshold range, the sampling point corresponding to the target parameter is identified as abnormal (i.e., noise) and filtered out. The target parameter can be any parameter among the equipment parameters and their corresponding upstream and downstream process parameters.

[0091] Principal component analysis (PCA) is a data dimensionality reduction technique. Hotelling's T2 threshold is a threshold used in PCA dimensionality reduction; however, since PCA is a readily available technology, it will not be elaborated upon here. When a principal component value exceeds the Hotelling's T2 threshold (e.g., the 95% threshold used in this example for hypothesis testing), the original sample data for that principal component is considered anomaly (i.e., noise) and is filtered out.

[0092] Interpolation is used for samples with missing data. Besides data gaps caused by the preprocessing methods mentioned above (parameter validity analysis and principal component analysis), missing data may also exist in historical sample data itself, thus requiring interpolation. However, in most cases, the missing data samples are relatively scattered, while in a few cases they are relatively concentrated. Therefore, appropriate interpolation processes need to be performed according to their distribution:

[0093] When the number of missing consecutive data is less than or equal to the first missing data threshold (e.g., 10 samples), data is supplemented by linear interpolation.

[0094] When the continuous data missing amount is greater than the first missing amount threshold but less than or equal to the second missing amount threshold (e.g., 50 samples), a preset number (e.g., 5 to 10 samples) of reference sample data on both sides of the data missing point are obtained, and data interpolation is performed to supplement the data based on the mean of the reference sample data, a bidirectional sliding data window, and linear proportional interpolation. For example, the mean of the reference sample data plus the bidirectional sliding data window is used to calculate the predicted value of each data point, and the final filling value is calculated using a linear proportional method.

[0095] When the amount of missing continuous data is greater than the second missing data threshold, the device or apparatus corresponding to the sampling point is determined to be abnormal due to the lack of large continuous sample data.

[0096] Correlation analysis does not rely on process experience (such as Bernoulli's equation) to establish a correlation. Instead, it analyzes the similarity between any two parameters in the original sample data. If the similarity is within a preset range (e.g., 0.95 to 1), the two parameters are considered correlated, and the data corresponding to one of the parameters is reduced.

[0097] Sample sparsity processing is used when the sample size is large, typically exceeding 10,000 or even reaching 100,000 samples. Therefore, it's necessary to perform sampling and filtering to create a new sample matrix without affecting the overall trend. For example, when the number of original sample data exceeds the maximum sample threshold, the original sample data is sampled and filtered according to a preset time interval.

[0098] It should be noted that, depending on the actual situation, the above preprocessing methods may not be performed in the same order (for example, interpolation may be performed first, followed by correlation analysis, then parameter validity analysis and principal component analysis), and the above preprocessing methods may not be run only once (i.e., if correlation analysis is performed after interpolation, followed by parameter validity analysis and principal component analysis, then interpolation may be performed again to complete the time series).

[0099] It should be noted that both parameter validity analysis and principal component analysis are noise reduction processes. In one embodiment, parameter validity analysis is used for preliminary noise reduction, and principal component analysis is used for further noise reduction.

[0100] By preprocessing, samples can be selectively screened, missing values ​​can be added, parameters can be reduced, and the sample size can be appropriately reduced. Users can perform further diagnostic and early warning operations based on a smaller amount of complete data, making fault diagnosis more convenient and faster.

[0101] Specifically, in the method described above, after performing empirical mode decomposition on the device parameters, the method further includes:

[0102] Based on the number of IMF components obtained from the decomposition of the equipment parameters, determine the maximum reference number of IMF components;

[0103] IMF components are supplemented based on the maximum reference quantity, wherein the value of the supplemented IMF component is 0.

[0104] In this embodiment, the operations following the empirical mode decomposition of the device parameters are further explained. Since the number of IMF components (i.e., the dimensions of both the IMF components and the device covariance matrix) must be equal during calculations, a dimensional comparison is performed between the number of IMF components obtained from the device parameter decomposition and the pre-acquired device covariance matrix. If the current number of IMF components is greater than or equal to the number of dimensions of the device covariance matrix, then the current number of IMF components is determined as the maximum reference number of IMF components. Otherwise, the number of dimensions of the device covariance matrix is ​​determined as the maximum reference number of IMF components. Then, based on this maximum reference number, the dimensions of the current IMF components or the device covariance matrix are supplemented with IMF components, where the value of the supplemented IMF components is 0.

[0105] Furthermore, the method described above also includes:

[0106] After determining the operating status, the sample data to be analyzed is displayed and stored according to the operating status and / or the target time point, and the operating status is output.

[0107] In this embodiment, after determining the operating status of the target device, to facilitate the user's understanding of the target device's operating status, the sample data to be analyzed will be displayed and stored based on the operating status and / or the target time point, and the operating status will be output. Displaying and storing the sample data to be analyzed allows the user to understand the device's operating data at the target time point in detail, and storing it facilitates subsequent querying or calculation as historical data.

[0108] Specifically, in the case where the operating state is the abnormal state, the method described above further includes:

[0109] An early warning is issued based on the first robust distance and / or the second robust distance used to determine the abnormal state.

[0110] In this embodiment, if the operating state is determined to be abnormal, an early warning device will also issue a warning, such as by changing the color and frequency of lights, emitting sounds, or displaying text. This allows the user to take timely action based on the warning. Specifically, the warning can be differentiated based on the cause of the abnormal state, such as an abnormal first robust distance and / or an abnormal second robust distance, allowing the user to take targeted actions.

[0111] Specifically, if the second robust distance is abnormal, it is considered that the upstream and downstream processes of the target equipment are abnormal, and this warning information can be provided to the process module for process engineers to view and make decisions; if the first robust distance condition is abnormal, it is considered that the target equipment is abnormal, and this information can be provided to the equipment module for equipment engineers to view and make decisions.

[0112] To facilitate understanding by those skilled in the art, the following description uses a specific example of fault diagnosis and monitoring (including related verification) of a quench tower circulating pump (target equipment).

[0113] For faults caused by impeller damage in the quench tower circulating pump, a total of 61,820 samples were selected for diagnosis from April 20, 2022 to June 1, 2022. A total of 23 parameters were selected, among which the equipment parameter of the quench tower circulating pump was current. Pre-processing details are shown in Table 1.

[0114] Table 1 Preprocessing Details

[0115]

[0116] The preprocessed current data is then visualized, such as... Figure 3 As shown, it is clear that the current values ​​between samples 670 and 970 are extremely small, which is actually due to maintenance. In addition, the current is generally decreasing, but it has not actually reached the warning value, so no alarm is triggered. Therefore, it is difficult to identify the fault situation and further processing is required.

[0117] Using the above 12 parameters and 1032 samples to calculate the robust distance and the corresponding threshold (where p = 12), the results are as follows: Figure 4 As shown in the figure, the abnormal samples are concentrated in the first 300 samples, corresponding to samples from April 20th to April 28th. Further verification through process analysis revealed significant fluctuations in all pressure and flow control parameters in these samples. Simultaneously, the fault detection rate was calculated to be 83.87%, which is relatively high. Therefore, the fault originated from an abnormal process condition; specifically, the fault was caused by impeller damage due to silicon powder inclusions in the upstream fluid.

[0118] However, for samples between 670 and 970, the current became extremely small due to the pump replacement, but in Figure 4 The current parameter was not detected. Therefore, it is also necessary to perform diagnostic detection for the equipment parameter. This is why the separate analysis of equipment parameters was included in the above steps.

[0119] For the current samples, no missing samples were found, so empirical mode decomposition was performed directly, and the results are as follows. Figure 5As shown, six IMF components and one residual component were obtained. It is evident that after EMD decomposition, all components exhibit significant fluctuations between the 670th and 970th samples. Next, using these seven components, their robust distances were calculated, and the results are as follows... Figure 6 As shown, significant fluctuations can also be observed within the same time period, and the fluctuations are even greater. If pump replacement is considered a fault, the calculated fault detection rate is 100%.

[0120] Based on the two robust distance samples mentioned above, abnormal process conditions and equipment malfunctions can be clearly observed.

[0121] Then, using the obtained covariance matrix, mean matrix, and threshold as a benchmark, the real-time updated data is substituted to obtain the real-time updated robust distance sample, and it is determined whether it exceeds the threshold to trigger an alarm. On July 15th, it was found that the robust distance of the entire parameter set (the set of equipment parameters and their upstream and downstream process parameters as a whole) exceeded the threshold, while the robust distance of the equipment parameter component set did not exceed the threshold, indicating that there was an anomaly in the upstream and downstream processes. Upon inspection, it was found that the upstream flow rate surged, while the downstream flow rate briefly decreased. Further inspection revealed that the fluid contained a large amount of silicon powder, thus requiring adjustments to the upstream flow rate and the cyclone separator in the fluidized bed.

[0122] See Figure 7 Another embodiment of this application also provides a fault diagnosis device for polysilicon production equipment, comprising:

[0123] The first processing module 701 is used to acquire the sample data to be analyzed collected by the target device at the target time point and the first historical sample data collected at multiple first historical time points respectively. The time difference between the first historical time point and the target time point is less than a preset time difference. The target time is the monitoring time point or any historical time point.

[0124] The second processing module 702 is used to determine, based on the sample data to be analyzed and the first historical sample data, a first set consisting of preset equipment parameters and a second set consisting of the equipment parameters and their corresponding upstream and downstream process parameters.

[0125] The third processing module 703 is used to perform empirical mode decomposition on the device parameters in the first set to obtain a third set.

[0126] The fourth processing module 704 is configured to determine a first robust distance of the equipment parameters based on the third set, the equipment mean vector, and the equipment covariance matrix; and to determine a second robust distance of the equipment parameters and their corresponding upstream and downstream process parameters based on the second set, the overall mean vector, and the overall covariance matrix, wherein the equipment mean vector and the equipment covariance matrix are pre-determined based on the equipment parameters in multiple second historical data sets, and the overall mean vector and the overall covariance matrix are pre-determined based on the equipment parameters and their corresponding upstream and downstream process parameters in multiple second historical data sets.

[0127] The fifth processing module 705 is used to determine that the operating state of the target device within the sampling time corresponding to the target time point is an abnormal state if the first robust distance is greater than a predetermined first robust distance threshold and / or the second robust distance is greater than a predetermined second robust distance threshold; otherwise, it determines that the operating state is a normal state.

[0128] Preferably, the apparatus as described above further includes:

[0129] The sixth processing module is used to acquire multiple sets of the second historical sample data from the target device;

[0130] The seventh processing module is used to determine, based on the second historical sample data, a fourth set consisting of the equipment parameters and a fifth set consisting of the equipment parameters and their corresponding upstream and downstream process parameters.

[0131] The eighth processing module is used to perform empirical mode decomposition on the device parameters in the fourth set to obtain the sixth set;

[0132] The ninth processing module is used to determine the device mean vector, the device covariance matrix, and the third robust distance based on the sixth set; and to determine the overall mean vector, the overall covariance matrix, and the fourth robust distance based on the fifth set.

[0133] Furthermore, the apparatus described above also includes:

[0134] The tenth processing module is used to determine the first robust distance threshold based on the third robust distance, and to determine the second robust distance threshold based on the fourth robust distance.

[0135] Specifically, the device described above further includes:

[0136] The eleventh processing module is used to preprocess the original sample data according to a preset preprocessing strategy to obtain the preprocessed original sample data, wherein the original sample data is the sample data to be analyzed and the first historical sample data, or the second historical sample data.

[0137] The preprocessing includes at least one of the following:

[0138] Parameter validity analysis;

[0139] Principal component analysis;

[0140] Interpolation processing;

[0141] Correlation analysis;

[0142] Sample sparsity processing.

[0143] Specifically, the device described above further includes:

[0144] The twelfth processing module is used to determine the maximum reference number of IMF components based on the number of IMF components obtained from the decomposition of the device parameters.

[0145] The thirteenth processing module is used to supplement the IMF components according to the maximum reference quantity, wherein the value of the supplemented IMF component is 0.

[0146] Furthermore, the apparatus described above also includes:

[0147] The fourteenth processing module is used to, after determining the operating status, display and store the sample data to be analyzed according to the operating status and / or the target time point, and output the operating status.

[0148] Specifically, the device described above further includes:

[0149] The fifteenth processing module is used to issue an early warning based on the first robust distance and / or the second robust distance used to determine the abnormal state.

[0150] The device embodiment of this application is a device corresponding to the embodiment of the above method. All implementation means in the above method embodiment are applicable to the device embodiment and can achieve the same technical effect.

[0151] See Figure 8 In another embodiment of this application, the structure of a fault diagnosis device for polysilicon production equipment is described in detail, including:

[0152] The point storage module 801 is used to collect the parameters and information input on site, and store the information and name tags of each parameter in it;

[0153] The sample storage module 802 stores samples of parameters input on-site and defines missing data values ​​as "*".

[0154] The event setting and storage module 803 is used to input the limiting conditions, such as the thresholds involved in the various preprocessing methods mentioned above. Generally, these conditions are static conditions, usually a constant value, or the result of basic operations. However, there are also cases of dynamic thresholds. Based on its mathematical principles, programming can be performed in it. In this special case, in order to facilitate programming, it can support the import of external languages.

[0155] Preprocessing module 804, namely the eleventh processing module mentioned above, is linked to event setting and storage module S403 and includes:

[0156] The parameter validity analysis unit is used for preliminary noise reduction.

[0157] Principal component analysis unit, used for further screening, for example, includes a pre-set principal component dimensionality reduction algorithm and Hotelling T. 2 Module for calculating the Squared Prediction Error (SPE) test value;

[0158] The correlation analysis unit has a preset correlation analysis algorithm and displays two-dimensional correlation analysis results, allowing users to manually filter parameters based on the results.

[0159] The interpolation and sampling unit is responsible for completing the sample values ​​of each parameter and filtering them according to a custom time interval.

[0160] Both the diagnostic module 805 (including at least one of the fourth, fifth, ninth, and tenth modules mentioned above) and the early warning module 806 (such as the fifteenth module mentioned above) are pre-set with a fast minimum covariance determinant algorithm to calculate the covariance matrix and mean matrix of the optimized historical data, and calculate the robust distance and threshold based on them. In addition, it also counts fault point information. For each sample, abnormal points are recorded as 1 and normal points are recorded as 0. Specifically, for the early warning module S408, when the covariance matrix and mean matrix are determined using historical data, the historical data will no longer be used (i.e., historical data does not occupy the storage of the early warning module). Only the covariance matrix and mean matrix are used to solve the robust distance of the real-time updated sample.

[0161] The data display module 807 (as the fourteenth processing module mentioned above) displays all processing procedures and corresponding charts (such as a time series diagram of a certain parameter, and the time range of the time series diagram can be adjusted at any time). The alarm module 808 is mainly for the real-time updated data of the early warning module 806. If the threshold is exceeded, it will light up red, which is an alarm. Under normal circumstances, it will be green.

[0162] The case storage module 809 backs up all operation steps and the results of the processed data as cases (and can be updated in real time as new operations occur), and outputs the processed data samples to a large display interface in real time. In addition, the backup is used so that if a serious accident occurs, the user can retrieve the cases in the case storage module 80 for analysis to determine the cause of the accident.

[0163] Another embodiment of this application provides a server, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the polysilicon production equipment fault diagnosis method as described above.

[0164] Another embodiment of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the polysilicon production equipment fault diagnosis method described above.

[0165] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0166] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0167] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for diagnosing faults in polysilicon production equipment, characterized in that, include: The target device acquires the sample data to be analyzed collected at the target time point and the first historical sample data collected at multiple first historical time points respectively. The time difference between the first historical time point and the target time point is less than a preset time difference. The target time is the monitoring time point or any historical time point. Based on the sample data to be analyzed and the first historical sample data, a first set consisting of preset equipment parameters and a second set consisting of the equipment parameters and their corresponding upstream and downstream process parameters are determined. Empirical mode decomposition is performed on the device parameters in the first set to obtain a third set; Based on the third set, the equipment mean vector, and the equipment covariance matrix, a first robust distance of the equipment parameters is determined; and based on the second set, the overall mean vector, and the overall covariance matrix, a second robust distance of the equipment parameters and their corresponding upstream and downstream process parameters is determined, wherein the equipment mean vector and the equipment covariance matrix are pre-determined based on the equipment parameters in multiple second historical sample data, and the overall mean vector and the overall covariance matrix are pre-determined based on the equipment parameters and their corresponding upstream and downstream process parameters in multiple second historical sample data; If the first robust distance is greater than a predetermined first robust distance threshold, and / or the second robust distance is greater than a predetermined second robust distance threshold, the operating state of the target device within the sampling time corresponding to the target time point is determined to be an abnormal state; otherwise, the operating state is determined to be a normal state.

2. The method according to claim 1, characterized in that, Also includes: Obtain multiple sets of the second historical sample data from the target device; Based on the second historical sample data, a fourth set consisting of the equipment parameters and a fifth set consisting of the equipment parameters and their corresponding upstream and downstream process parameters are determined. Empirical mode decomposition is performed on the device parameters in the fourth set to obtain the sixth set; Based on the sixth set, the device mean vector, the device covariance matrix, and the third robust distance are determined; and based on the fifth set, the overall mean vector, the overall covariance matrix, and the fourth robust distance are determined.

3. The method according to claim 2, characterized in that, Also includes: The first robust distance threshold is determined based on the third robust distance, and the second robust distance threshold is determined based on the fourth robust distance.

4. The method according to claim 1 or 2, characterized in that, After obtaining the original sample data, the method further includes: According to a preset preprocessing strategy, the original sample data is preprocessed to obtain the preprocessed original sample data, wherein the original sample data is the sample data to be analyzed and the first historical sample data, or the second historical sample data; The preprocessing includes at least one of the following: Parameter validity analysis; Principal component analysis; Interpolation processing; Correlation analysis; Sample sparsity processing.

5. The method according to claim 1 or 2, characterized in that, After performing empirical mode decomposition on the device parameters, the method further includes: Based on the number of intrinsic mode function (IMF) components obtained from the decomposition of the device parameters, determine the maximum reference number of IMF components; IMF components are supplemented based on the maximum reference quantity, wherein the value of the supplemented IMF component is 0.

6. The method according to claim 1, characterized in that, Also includes: After determining the operating status, the sample data to be analyzed is displayed and stored according to the operating status and / or the target time point, and the operating status is output.

7. The method according to claim 1 or 6, characterized in that, When the operating state is the abnormal state, the method further includes: An early warning is issued based on the first robust distance and / or the second robust distance used to determine the abnormal state.

8. A fault diagnosis device for polysilicon production equipment, characterized in that, include: The first processing module is used to acquire the sample data to be analyzed collected by the target device at the target time point and the first historical sample data collected at multiple first historical time points respectively. The time difference between the first historical time point and the target time point is less than a preset time difference. The target time is the monitoring time point or any historical time point. The second processing module is used to determine, based on the sample data to be analyzed and the first historical sample data, a first set consisting of preset equipment parameters and a second set consisting of the equipment parameters and their corresponding upstream and downstream process parameters. The third processing module is used to perform empirical mode decomposition on the device parameters in the first set to obtain a third set. The fourth processing module is configured to determine a first robust distance of the equipment parameters based on the third set, the equipment mean vector, and the equipment covariance matrix; and to determine a second robust distance of the equipment parameters and their corresponding upstream and downstream process parameters based on the second set, the overall mean vector, and the overall covariance matrix, wherein the equipment mean vector and the equipment covariance matrix are pre-determined based on the equipment parameters in multiple second historical sample data, and the overall mean vector and the overall covariance matrix are pre-determined based on the equipment parameters and their corresponding upstream and downstream process parameters in multiple second historical sample data. The fifth processing module is used to determine that the operating state of the target device within the sampling time corresponding to the target time point is an abnormal state if the first robust distance is greater than a predetermined first robust distance threshold and / or the second robust distance is greater than a predetermined second robust distance threshold; otherwise, it determines that the operating state is a normal state.

9. A server, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for diagnosing faults in polysilicon production equipment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the polysilicon production equipment fault diagnosis method as described in any one of claims 1 to 7.

Images