Membrane filament breaking point detection system and method for wet membrane dialyzer

Abstract

The invention belongs to the technical field of wet membrane dialyzer detection, and discloses a membrane yarn breaking point detection system and method for a wet membrane dialyzer. The system comprises a leakage detection positioning module, a membrane rupture feature extraction module, a rupture point type intelligent diagnosis module and a root cause correlation analysis module, analyzes a fusion data set, calculates the spatial position of a leakage point according to an analysis result to obtain a leakage event report, performs quantitative feature extraction based on the leakage event report, and performs root cause correlation analysis based on the quantitative feature extraction. The method comprises the following steps: obtaining a broken point feature vector, inputting the broken point feature vector into a pre-trained machine learning model, automatically identifying the root type of a broken point, obtaining a broken point type diagnosis report, positioning a production link reason based on the broken point type diagnosis report in combination with production batch data, and obtaining a root reason analysis report. The method has the remarkable advantages of being high in detection accuracy, high in root cause tracing intelligence and good in preventive protection effect.

Description

Technical Field

[0001] This invention relates to the field of wet membrane dialyzer testing technology, and more specifically, to a system and method for detecting membrane filament rupture in wet membrane dialyzers. Background Technology

[0002] A wet membrane dialyzer is a medical consumable used for hemodialysis treatment. Its core feature is that during the production process, hollow fiber membranes are soaked in a filling fluid such as sterile water or physiological saline, so that the membranes are always in a moist and expanded state. The working principle is based on semi-permeable membrane technology. The patient's blood flows through the inside of the membrane and the permeate flows on the outside of the membrane. Through diffusion, convection and ultrafiltration, toxins and excess water in the body are removed.

[0003] However, traditional wet membrane dialyzer membrane fiber breakage detection systems suffer from several shortcomings during use. First, these systems largely rely on operator visual inspection to pinpoint leaks. This manual method is not only inefficient but also susceptible to operator fatigue, environmental interference, and working hours, leading to missed detections and false alarms. Second, traditional systems often lack the ability to identify breakage types, requiring manual investigation and judgment of the cause. Furthermore, the judgment relies on human experience, resulting in low efficiency in tracing production factors and significant time commitment. Thirdly, the inspection and production processes of traditional systems are mostly isolated "information silos." Data such as the location and type of membrane rupture obtained during the inspection process are only used to determine whether a single product is qualified. There is a lack of in-depth information mining and production-end linkage, making it difficult for managers to identify the key reason for the high membrane rupture rate. This leads to the recurrence of similar problems. In general, how to effectively solve the problems of high false alarm rate, reliance on manual analysis of rupture causes, and difficulty in preventive protection in traditional systems has become the issue that current wet membrane dialyzer membrane fiber rupture detection systems need to address and solve.

[0004] In view of this, the present invention proposes a membrane filament breakage detection system for wet membrane dialyzers to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: A wet membrane dialyzer membrane filament rupture detection system, comprising: a wet membrane dialyzer tracer detection device, a leak detection and location module, a rupture feature extraction module, a rupture type intelligent diagnosis module, and a root cause correlation analysis module, wherein: A tracer detection device for wet membrane dialyzers is used to connect one end of the dialyzer (detected by water testing) to a tobacco smoke source, and the other end (the permeate) to a negative pressure vacuum pump. The negative pressure vacuum pump creates an airflow from the blood port to the permeate inside the dialyzer. Nicotine particles produced by tobacco combustion are used as tracers. When a rupture point exists in the dialyzer, a high concentration of smoke will be concentrated and leaked from the rupture point under the negative pressure. Then, the ruptured dialyzer is removed, and images of the parts at the rupture point are acquired. The leak detection and location module is used to analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results, and obtain a leak event report. The membrane rupture feature extraction module is used to perform quantitative feature extraction based on leakage event reports to obtain rupture point feature vectors. The intelligent diagnostic module for breakpoint types is used to input the feature vector of a breakpoint into a pre-trained machine learning model to automatically identify the fundamental type of the breakpoint and obtain a breakpoint type diagnostic report. The Root Cause Correlation Analysis module is used to locate the cause in the production process based on the failure point type diagnosis report and production batch data, and obtain the root cause analysis report.

[0006] The system also includes: a multi-channel sensor collaborative acquisition module, a data fusion preprocessing module, and a production parameter adaptive instruction generation module. The multi-channel sensor collaborative acquisition module is used to uniformly schedule the start-up and shutdown of the sensor network and the parameters, and to acquire raw data to obtain the raw metadata dataset. The data fusion preprocessing module is used to preprocess and fuse the original metadata to obtain a fused dataset. The fused dataset includes at least the air pressure data during the detection process, images of the parts at the fault point, and production information of the parts at the fault point. The production parameter adaptive instruction generation module is used to perform a comprehensive diagnosis of the root cause analysis report and obtain production parameter adjustment instructions.

[0007] Furthermore, the steps of uniformly scheduling the start-up and shutdown of the sensor network and its parameters, and collecting raw data, include: S1.1: The user starts the detection task based on the UI interface and receives the start command; S1.2: Monitor the start command in step S1.1. When the monitoring result is yes, generate a synchronous acquisition command set according to the preset detection scheme. The synchronous acquisition command set includes initialization command for all sensor channels, sampling rate alignment command for all sensor channels, sensor acquisition time synchronization command for all channels, and stop acquisition condition command. S1.3: Output the synchronous acquisition instruction set to the sensor network and perform data acquisition to obtain the raw dataset; S1.4: Pack the sensor ID, sampling rate, measurement range, and raw dataset to obtain the raw meta-dataset; S1.5: Output the original metadata to the data fusion preprocessing module.

[0008] The data fusion preprocessing module is used to preprocess and fuse the original metadata to obtain a fused dataset; Furthermore, the steps for preprocessing and fusing the original metadata include: S2.1: Read the original dataset from the original metadata set, and interpolate the original dataset according to the time synchronization instructions of all channel sensors in step S1.2 to obtain the time-aligned dataset; S2.2: A zero-phase low-pass filter is used to digitally filter the time-aligned dataset to obtain a denoised dataset. The specific calculation formula for the zero-phase low-pass filter is as follows: ; Get the time point Filtered data ,in, This is a zero-phase digital filter function. These are the parameters of the low-pass filter. For time points Time-aligned datasets; S2.3: Based on the measurement range in the original metadata set, the data denoising dataset is normalized to obtain a normalized dataset. The specific formula for normalization is: ; Get the first Each sensor channel at time point Normalized data ,in, For the first Each sensor channel time point Filtered data, and For the first Minimum and maximum values ​​of each sensor channel within the measurement range; S2.4: By time point Based on this, the normalized data from all sensor channels are aggregated into a single feature vector, and the feature vectors from all time points within the detection period are integrated to obtain a fused dataset. S2.5: Output the fused dataset to the leak detection and localization module.

[0009] The leak detection and location module is used to analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results, thereby generating a leak event report.

[0010] Further, the steps of analyzing the fused dataset and calculating the spatial location of the leak point based on the analysis results include: S3.1: Based on the fused dataset, calculate the data energy of all sensor channels and compare it with the leakage judgment threshold. When the data energy is greater than the leakage judgment threshold, generate a leakage signal and proceed to step S3.2. The specific calculation formula for data energy is as follows: ; Get the time point Data energy ,in, This refers to the number of sensor channels; S3.2: When a leakage signal is detected, the leakage trigger timestamp data is analyzed using a signal strength-based weighted centroid localization algorithm to obtain the leakage point coordinates. The specific formula for the signal strength-based weighted centroid localization algorithm is as follows: ; Obtain the coordinates of the leak point ,in, For the first Weighting factors for each sensor channel For the first The position coordinates of each sensor channel; S3.3: Pack the sensor ID of the leaking sensor, the coordinates of the leak point, the maximum data energy, the leak trigger timestamp, and the preset leak time window to obtain a leak event report; S3.4: Output the leakage event report to the membrane rupture feature extraction module; The membrane rupture feature extraction module is used to extract quantitative features based on leakage event reports to obtain rupture point feature vectors.

[0011] Furthermore, the steps for quantitative feature extraction based on the leak incident report include: S4.1: Based on the preset leakage time window in the leakage event report, the feature vectors in the dataset corresponding to the time points of the preset leakage time window are fused in advance to obtain the truncated dataset; S4.2: Perform statistical calculations on the time-domain features in the extracted dataset to obtain a time-domain feature vector. The time-domain feature vector includes peak rise time feature, peak duration feature, and data distribution skewness feature, where: The peak rise time characteristic is obtained by statistically extracting the time required for the peak to rise from 10% to 90% in the dataset; The peak duration feature is obtained by statistically extracting the duration of more than 50% of the peaks in the dataset; The skewness characteristic of the data distribution is determined by substituting it into the calculation formula: We obtained, among which, To extract the number of feature vectors in the dataset, For the first 1 eigenvector To extract the mean of the feature vectors in the dataset, To extract the standard deviation of the feature vectors in the dataset; S4.3: Utilize the Fast Fourier Transform to calculate the frequency domain features in the truncated dataset, obtaining a frequency domain feature vector. This vector includes centroid frequency features and bandwidth features, where: The centroid frequency characteristic is the centroid frequency of the spectrum; The bandwidth feature is the bandwidth of the feature vector; S4.4: Combine the time-domain feature vector and the frequency-domain feature vector with a preset fixed length to obtain the breaking point feature vector; S4.5: Output the feature vector of the break point to the intelligent diagnosis module for the break point type; The intelligent diagnostic module for breakpoint types is used to input the feature vector of a breakpoint into a pre-trained machine learning model to automatically identify the fundamental type of the breakpoint and obtain a breakpoint type diagnostic report.

[0012] Furthermore, the steps for automatically identifying the fundamental type of the breach point by inputting the feature vector of the breach point into a pre-trained machine learning model include: S5.1: Load the pre-trained machine learning model; S5.2: Input the feature vector of the break point into the machine learning model, and output the predicted break point type and the confidence level of the prediction result; S5.3: Obtain a break point type diagnostic report by taking the packaging production equipment ID, predicted break point type, prediction result confidence level, and break point feature vector; S5.4: Output the failure point type diagnostic report to the root cause association analysis module.

[0013] The Root Cause Correlation Analysis module is used to locate the cause in the production process based on the failure point type diagnosis report and production batch data, and obtain the root cause analysis report.

[0014] Furthermore, based on the defect type diagnostic report and production batch data, the steps for locating the cause in the production process include: S6.1: Based on the database, query the relevant information of the current production batch to obtain the data for generating the current production batch. The relevant information includes material batch information, production equipment ID information, breakage type and operator ID information. S6.2: Based on the predicted break point type in the break point type diagnosis report, query the database for historical production batch data that matches the break point type to obtain the historical production batch dataset; S6.3: Match the predicted breach type in the breach type diagnosis report with the current production batch data, and perform root cause matching based on the historical production batch dataset in step S6.2 to obtain the leak root cause report; S6.4: Package the root cause report of the leak, the overlap, and the production parameter logs containing the production equipment ID information in the root cause report of the leak to obtain the root cause analysis report; S6.5: Output the root cause analysis report to the production parameter adaptive instruction generation module.

[0015] The production parameter adaptive instruction generation module is used to perform a comprehensive diagnosis of the root cause analysis report and obtain production parameter adjustment instructions. Furthermore, the steps for a comprehensive diagnosis based on the root cause analysis report include: S7.1: Match the fault type diagnosis report and root cause analysis report with the preset adjustment strategy to obtain production parameter adjustment instructions; S7.2: Convert production parameter adjustment instructions into execution instruction format and output them to the intelligent controller; S7.3: Record the instruction issuance status and instruction execution feedback in step S7.2 to obtain an adjustment operation report; S7.4: Store the adjustment operation report in the database.

[0016] Furthermore, a method for detecting membrane filament breakage in a wet membrane dialyzer is also provided, comprising the following steps according to any of the aforementioned wet membrane dialyzer membrane filament breakage detection systems: S1: Perform unified scheduling of the start-up, shutdown, and parameters of the sensor network, and collect raw data to obtain the raw metadata dataset; S2: Preprocess and fuse the original metadata to obtain the fused dataset; S3: Analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results to obtain a leak event report; S4: Quantitative feature extraction is performed based on the leak event report to obtain the feature vector of the breach point; S5: Input the feature vector of the breach point into the pre-trained machine learning model to automatically identify the fundamental type of the breach point and obtain a breach point type diagnosis report; S6: Based on the failure point type diagnostic report and production batch data, locate the cause in the production process and obtain the root cause analysis report; S7: Conduct a comprehensive diagnosis based on the root cause analysis report and obtain instructions for adjusting production parameters.

[0017] The technical effects and advantages of the wet membrane dialyzer membrane filament breakage detection system of the present invention are as follows: This invention achieves unified scheduling of sensor network startup, shutdown, and parameter control, and collects raw data to obtain raw metadata. The raw metadata is then preprocessed and fused to obtain a fused dataset. This fused dataset is analyzed, and the spatial location of leak points is calculated based on the analysis results to generate a leak event report. Based on the leak event report, quantitative feature extraction is performed to obtain a breach feature vector. This breach feature vector is input into a pre-trained machine learning model to automatically identify the root type of the breach, generating a breach type diagnostic report. Based on the breach type diagnostic report and production batch data, the cause in the production process is located, generating a root cause analysis report. A comprehensive diagnosis is performed on the root cause analysis report to generate production parameter adjustment instructions. This allows the system to work collaboratively through a multi-channel sensor collaborative acquisition module, a data fusion preprocessing module, and a leak detection and location module. It replaces the traditional system's visual observation with highly efficient and multi-dimensional sensor data, and performs real-time analysis of the collected raw metadata, significantly reducing the occurrence of missed detections and false alarms in traditional systems. In addition, the present invention quantifies leakage signals by combining a membrane rupture feature extraction module with a rupture point type intelligent diagnosis module, and automatically classifies them through a machine learning model, which greatly reduces the time required for manual analysis in traditional systems. Finally, the root cause correlation analysis module deeply explores the root causes of quality problems, and combines this with the production parameter adaptive instruction generation module to automatically generate precise adjustment instructions, thereby reducing the production defect rate from the source. Overall, this invention has significant advantages such as high detection accuracy, strong intelligent root cause tracing, and good preventive protection. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a membrane filament breakage detection system for a wet membrane dialyzer according to the present invention; Figure 2 This is a schematic diagram of the process for detecting membrane filament breakage in a wet membrane dialyzer according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0021] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0022] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0023] In practice, the server-side equipment deployed in the wet membrane dialyzer membrane filament breakage detection system may consist of one or more devices. This system can be implemented as a business instance, a virtual machine, or a hardware device. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node to provide membrane filament breakage detection to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage each user terminal. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more devices configured to provide the system to various user terminals.

[0024] In terms of implementation, the wet membrane dialyzer membrane filament breakage detection system and the user terminal are mutually compatible. That is, if the wet membrane dialyzer membrane filament breakage detection system is implemented as an application installed on a cloud service platform, then the user terminal is a client that establishes a communication connection with the application; or if the wet membrane dialyzer membrane filament breakage detection system is implemented as a website, then the user terminal is implemented as a webpage; or if the wet membrane dialyzer membrane filament breakage detection system is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0025] like Figure 1 The figure shown is a system architecture diagram of a wet membrane dialyzer membrane filament breakage detection system provided in an embodiment of the present invention.

[0026] The wet membrane dialyzer membrane filament breakage detection system of this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the wet membrane dialyzer membrane filament breakage detection system may include a multi-channel sensor collaborative acquisition module, a data fusion preprocessing module, a leak detection and location module, a membrane rupture feature extraction module, a breakage type intelligent diagnosis module, a root cause correlation analysis module, and a production parameter adaptive instruction generation module. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0027] In this embodiment of the invention, in the wet membrane dialyzer membrane filament breakage detection system, each of the above-mentioned modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. For example, the sharing and evaluation module can call the same information acquisition module to obtain information collected by that module. Based on the above characteristics, in the wet membrane dialyzer membrane filament breakage detection system provided in this embodiment of the invention, without modifying the program code, the applicable scope of the wet membrane dialyzer membrane filament breakage detection system architecture can be adjusted by adding modules and directly calling them, achieving cluster-based horizontal expansion to quickly and flexibly expand the wet membrane dialyzer membrane filament breakage detection system. In practical applications, the above modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0028] Example 1 Please see Figure 1 As shown, this embodiment provides a wet membrane dialyzer membrane fiber rupture detection system. The system includes: a wet membrane dialyzer tracer detection device, a leak detection and location module, a rupture feature extraction module, a rupture type intelligent diagnosis module, and a root cause correlation analysis module, wherein: A tracer detection device for wet membrane dialyzers is used to connect one end of the dialyzer (detected by water testing) to a tobacco smoke source, and the other end (the permeate) to a negative pressure vacuum pump. The negative pressure vacuum pump creates an airflow from the blood port to the permeate inside the dialyzer. Nicotine particles produced by tobacco combustion are used as tracers. When a rupture point exists in the dialyzer, a high concentration of smoke will be concentrated and leaked from the rupture point under the negative pressure. Then, the ruptured dialyzer is removed, and images of the parts at the rupture point are acquired. The leak detection and location module is used to analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results, and obtain a leak event report. The membrane rupture feature extraction module is used to perform quantitative feature extraction based on leakage event reports to obtain rupture point feature vectors. The intelligent diagnostic module for breakpoint types is used to input the feature vector of a breakpoint into a pre-trained machine learning model to automatically identify the fundamental type of the breakpoint and obtain a breakpoint type diagnostic report. The Root Cause Correlation Analysis module is used to locate the cause in the production process based on the failure point type diagnosis report and production batch data, and obtain the root cause analysis report.

[0029] The system also includes: a multi-channel sensor collaborative acquisition module, a data fusion preprocessing module, and a production parameter adaptive instruction generation module. The multi-channel sensor collaborative acquisition module is used to uniformly schedule the start-up and shutdown of the sensor network and the parameters, and to acquire raw data to obtain the raw metadata dataset. The data fusion preprocessing module is used to preprocess and fuse the original metadata to obtain a fused dataset. The fused dataset includes at least the air pressure data during the detection process, images of the parts at the fault point, and production information of the parts at the fault point. The production parameter adaptive instruction generation module is used to perform a comprehensive diagnosis of the root cause analysis report and obtain production parameter adjustment instructions.

[0030] Furthermore, the steps of uniformly scheduling the start-up and shutdown of the sensor network and its parameters, and collecting raw data, include: S1.1: The user starts the detection task based on the UI interface and receives the start command; S1.2: Monitor the start command in step S1.1. When the monitoring result is yes, generate a synchronous acquisition command set according to the preset detection scheme. The synchronous acquisition command set includes initialization command for all sensor channels, sampling rate alignment command for all sensor channels, sensor acquisition time synchronization command for all channels, and stop acquisition condition command. It should be explained that the stop acquisition condition command refers to, for example, reaching a preset acquisition command or receiving a stop signal from another module; S1.3: Output the synchronous acquisition instruction set to the sensor network and perform data acquisition to obtain the raw dataset; It should be explained that the original dataset includes, but is not limited to, air pressure data during the detection process, images of parts at the breach point, and production information of parts at the breach point; S1.4: Pack the sensor ID, sampling rate, measurement range, and raw dataset to obtain the raw meta-dataset; S1.5: Output the original metadata to the data fusion preprocessing module; The data fusion preprocessing module is used to preprocess and fuse the original metadata to obtain a fused dataset.

[0031] Further steps for preprocessing and fusing the original metadata include: S2.1: Read the original dataset from the original metadata set, and interpolate the original dataset according to the time synchronization instructions of all channel sensors in step S1.2 to obtain the time-aligned dataset; S2.2: A zero-phase low-pass filter is used to digitally filter the time-aligned dataset to obtain a denoised dataset. The specific calculation formula for the zero-phase low-pass filter is as follows: ; Get the time point Filtered data ,in, This is a zero-phase digital filter function. These are the parameters of the low-pass filter. For time points Time-aligned datasets; S2.3: Based on the measurement range in the original metadata set, the data denoising dataset is normalized to obtain a normalized dataset. The specific formula for normalization is: ; Get the first Each sensor channel at time point Normalized data ,in, For the first Each sensor channel time point Filtered data, and For the first Minimum and maximum values ​​of each sensor channel within the measurement range; S2.4: By time point Based on this, the normalized data from all sensor channels are aggregated into a single feature vector, and the feature vectors from all time points within the detection period are integrated to obtain a fused dataset. S2.5: Output the fused dataset to the leak detection and localization module.

[0032] The leak detection and location module is used to analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results, thereby generating a leak event report.

[0033] Further steps include analyzing the fused dataset and calculating the spatial location of the leak point based on the analysis results: S3.1: Based on the fused dataset, calculate the data energy of all sensor channels and compare it with the leakage judgment threshold. When the data energy is greater than the leakage judgment threshold, generate a leakage signal and proceed to step S3.2. The specific calculation formula for data energy is as follows: ; Get the time point Data energy ,in, This refers to the number of sensor channels; S3.2: When a leakage signal is detected, the leakage trigger timestamp data is analyzed using a signal strength-based weighted centroid localization algorithm to obtain the leakage point coordinates. The specific formula for the signal strength-based weighted centroid localization algorithm is as follows: ; Obtain the coordinates of the leak point ,in, For the first Weighting factors for each sensor channel For the first The position coordinates of each sensor channel; S3.3: Pack the sensor ID of the leaking sensor, the coordinates of the leak point, the maximum data energy, the leak trigger timestamp, and the preset leak time window to obtain a leak event report; It should be explained that the preset leakage time window refers to a period of time before and after the leakage trigger timestamp, such as 30 seconds before and after the leakage trigger timestamp; S3.4: Output the leakage event report to the membrane rupture feature extraction module.

[0034] The membrane rupture feature extraction module is used to perform quantitative feature extraction based on the leakage event report to obtain the rupture point feature vector.

[0035] Furthermore, the steps for quantitative feature extraction based on the leak incident report include: S4.1: Based on the preset leakage time window in the leakage event report, the feature vectors in the dataset corresponding to the time points of the preset leakage time window are fused in advance to obtain the truncated dataset; S4.2: Perform statistical calculations on the time-domain features in the extracted dataset to obtain a time-domain feature vector. The time-domain feature vector includes peak rise time feature, peak duration feature, and data distribution skewness feature, where: It should be explained that the peak value refers to the maximum value of a feature vector in the dataset. The value of that feature vector is then considered the peak value. The peak rise time characteristic is obtained by statistically extracting the time required for the peak to rise from 10% to 90% in the dataset; The peak duration feature is obtained by statistically extracting the duration of more than 50% of the peaks in the dataset; The skewness characteristic of the data distribution is determined by substituting it into the calculation formula: We obtained, among which, To extract the number of feature vectors in the dataset, For the first 1 eigenvector To extract the mean of the feature vectors in the dataset, To extract the standard deviation of the feature vectors in the dataset; S4.3: Utilize the Fast Fourier Transform to calculate the frequency domain features in the truncated dataset, obtaining a frequency domain feature vector. This vector includes centroid frequency features and bandwidth features, where: The centroid frequency characteristic is the centroid frequency of the spectrum; The bandwidth feature is the bandwidth of the feature vector; S4.4: Combine the time-domain feature vector and the frequency-domain feature vector with a preset fixed length to obtain the breaking point feature vector; S4.5: Output the feature vector of the break point to the intelligent diagnosis module for the break point type.

[0036] The intelligent diagnostic module for the type of breach is used to input the feature vector of the breach into a pre-trained machine learning model to automatically identify the fundamental type of the breach and obtain a breach type diagnostic report.

[0037] Furthermore, the steps for automatically identifying the fundamental type of the breach point by inputting the feature vector of the breach point into a pre-trained machine learning model include: S5.1: Load the pre-trained machine learning model; It should be explained that the machine learning model is, for example, a neural network model or a support vector machine, used to classify the feature vectors of the breakpoints; S5.2: Input the feature vector of the break point into the machine learning model, and output the predicted break point type and the confidence level of the prediction result; It should be explained that the types of breakage points include, but are not limited to, fiber breakage, adhesive problems, and foreign body punctures; the confidence level of the prediction result is the confidence level of the prediction result, which is a probability value between [0,1]. S5.3: Obtain a break point type diagnostic report by taking the packaging production equipment ID, predicted break point type, prediction result confidence level, and break point feature vector; S5.4: Output the failure point type diagnostic report to the root cause association analysis module.

[0038] The root cause correlation analysis module is used to locate the cause in the production process based on the failure point type diagnosis report and production batch data, and obtain the root cause analysis report.

[0039] Furthermore, based on the defect type diagnostic report and production batch data, the steps to pinpoint the cause in the production process include: S6.1: Based on the database, query the relevant information of the current production batch to obtain the data for generating the current production batch. The relevant information includes material batch information, production equipment ID information, breakage type and operator ID information. S6.2: Based on the predicted break point type in the break point type diagnosis report, query the database for historical production batch data that matches the break point type to obtain the historical production batch dataset; S6.3: Match the predicted breach type in the breach type diagnosis report with the current production batch data, and perform root cause matching based on the historical production batch dataset in step S6.2 to obtain the leak root cause report; It should be explained that this step refers to using the predicted breach type in the breach type diagnosis report and the current production batch data as a benchmark, and matching the data items with the historical generated batch dataset in step S6.2. For example, if the content of a certain historical production batch data has the highest overlap with the predicted breach type in the breach type diagnosis report and the content of the current production batch data, then that historical production batch data is the root cause report of the leak. S6.4: Package the root cause report of the leak, the overlap, and the production parameter logs containing the production equipment ID information in the root cause report of the leak to obtain the root cause analysis report; S6.5: Output the root cause analysis report to the production parameter adaptive instruction generation module; The production parameter adaptive instruction generation module is used to perform a comprehensive diagnosis of the root cause analysis report and obtain production parameter adjustment instructions. Further steps in conducting a comprehensive diagnosis based on the root cause analysis report include: S7.1: Match the fault type diagnosis report and root cause analysis report with the preset adjustment strategy to obtain production parameter adjustment instructions; It needs to be explained that the production parameter adjustment instruction means, for example, when the predicted break point type is an adhesive problem and the root cause analysis report is that the adhesive batch P is unevenly applied, an instruction to increase the adhesive spraying time is generated and used as a production parameter adjustment instruction. The adhesive spraying time increase instruction contains a set of characters representing an increase of 10% in the adhesive spraying time. S7.2: Convert production parameter adjustment instructions into execution instruction format and output them to the intelligent controller; S7.3: Record the instruction issuance status and instruction execution feedback in step S7.2 to obtain an adjustment operation report; S7.4: Store the adjustment operation report in the database.

[0040] This embodiment unifies the start-up, shutdown, and parameter scheduling of the sensor network, collects raw data to obtain a raw metadata dataset, preprocesses and fuses the raw metadata dataset to obtain a fused dataset, analyzes the fused dataset, calculates the spatial location of the leak point based on the analysis results, generates a leak event report, extracts quantitative features based on the leak event report to obtain a breach feature vector, inputs the breach feature vector into a pre-trained machine learning model to automatically identify the root type of the breach, generates a breach type diagnosis report, combines the breach type diagnosis report with production batch data to locate the cause in the production process, generates a root cause analysis report, performs a comprehensive diagnosis based on the root cause analysis report, and generates production parameter adjustment instructions. This enables the system to utilize a multi-channel sensor collaborative acquisition module and a data fusion preprocessing module... The block and leak detection and location module work together to replace the traditional system's visual observation with highly efficient and multi-dimensional sensor data, and perform real-time analysis of the collected raw metadata, greatly reducing the occurrence of missed detections and false alarms in traditional systems. In addition, this invention also quantifies the leakage signal by combining the membrane rupture feature extraction module with the intelligent diagnosis module for rupture point type, and automatically classifies it through machine learning models, greatly reducing the time required for manual analysis in traditional systems. Finally, the root cause correlation analysis module deeply explores the root causes of quality problems, and analyzes them in conjunction with the production parameter adaptive instruction generation module to automatically generate precise adjustment instructions, thereby reducing the production defect rate from the source. Overall, this invention has significant advantages such as high detection accuracy, strong intelligent root cause tracing, and good preventive protection effects.

[0041] Example 2 Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A method for detecting membrane filament rupture points in a wet membrane dialyzer is provided, the method comprising: S1: Perform unified scheduling of the start-up, shutdown, and parameters of the sensor network, and collect raw data to obtain the raw metadata dataset; S2: Preprocess and fuse the original metadata to obtain the fused dataset; S3: Analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results to obtain a leak event report; S4: Quantitative feature extraction is performed based on the leak event report to obtain the feature vector of the breach point; S5: Input the feature vector of the breach point into the pre-trained machine learning model to automatically identify the fundamental type of the breach point and obtain a breach point type diagnosis report; S6: Based on the failure point type diagnostic report and production batch data, locate the cause in the production process and obtain the root cause analysis report; S7: Conduct a comprehensive diagnosis based on the root cause analysis report and obtain instructions for adjusting production parameters.

[0042] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the present invention.

Claims

1. A membrane filament breakage detection system for a wet membrane dialyzer, characterized in that, The system includes: a wet membrane dialyzer tracer detection device, a leak detection and location module, a membrane rupture feature extraction module, a rupture point type intelligent diagnosis module, and a root cause correlation analysis module, wherein: The wet membrane dialyzer tracer detection device is used to connect one end of the dialyzer (detected by water testing) to a tobacco smoke source and the other end (connected to a vacuum pump). The vacuum pump creates an airflow from the blood port to the through port inside the dialyzer. Nicotine particles produced by tobacco combustion are used as tracers. When a rupture point exists in the dialyzer, a high concentration of smoke will leak out from the rupture point under the drive of negative pressure. Then, the ruptured dialyzer is removed and an image of the parts at the rupture point is acquired. The leak detection and localization module is used to analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results, thereby generating a leak event report. The membrane rupture feature extraction module is used to perform quantitative feature extraction based on the leakage event report to obtain the rupture point feature vector; The intelligent diagnostic module for the type of breach is used to input the feature vector of the breach into a pre-trained machine learning model, automatically identify the fundamental type of the breach, and obtain a breach type diagnostic report. The root cause correlation analysis module is used to locate the cause in the production process based on the failure point type diagnosis report and production batch data, and obtain the root cause analysis report.

2. The wet membrane dialyzer membrane filament breakage detection system according to claim 1, characterized in that, The system also includes: a multi-channel sensor collaborative acquisition module, a data fusion preprocessing module, and a production parameter adaptive instruction generation module, wherein: The multi-channel sensor collaborative acquisition module is used to uniformly schedule the start-up and shutdown of the sensor network and the parameters, and to acquire raw data to obtain raw metadata. The data fusion preprocessing module is used to preprocess and fuse the original metadata to obtain a fused dataset; The production parameter adaptive instruction generation module is used to perform a comprehensive diagnosis of the root cause analysis report and obtain production parameter adjustment instructions.

3. The wet membrane dialyzer membrane filament breakage detection system according to claim 2, characterized in that, The steps for unified scheduling of sensor network startup, shutdown, and parameter control, as well as the acquisition of raw data, include: S1.1: The user starts the detection task based on the UI interface and receives the start command; S1.2: Monitor the start command in step S1.

1. When the monitoring result is yes, generate a synchronous acquisition command set according to the preset detection scheme. The synchronous acquisition command set includes initialization command for all sensor channels, sampling rate alignment command for all sensor channels, sensor acquisition time synchronization command for all channels, and stop acquisition condition command. S1.3: Output the synchronous acquisition instruction set to the sensor network and perform data acquisition to obtain the raw dataset; S1.4: Pack the sensor ID, sampling rate, measurement range, and raw dataset to obtain the raw meta-dataset; S1.5: Output the original metadata to the data fusion preprocessing module.

4. The wet membrane dialyzer membrane filament breakage detection system according to claim 2, characterized in that, The steps for preprocessing and fusing the original metadata include: S2.1: Read the original dataset from the original metadata set, and interpolate the original dataset according to the time synchronization instructions of all channel sensors in step S1.2 to obtain the time-aligned dataset; S2.2: Use zero-phase low-pass filtering to perform digital filtering on the time-aligned dataset to obtain a data denoising dataset; S2.3: Normalize the data denoising dataset based on the measurement range in the original metadata set to obtain a normalized dataset; S2.4: By time point Based on this, the normalized data from all sensor channels are aggregated into a single feature vector, and the feature vectors from all time points within the detection period are integrated to obtain a fused dataset. S2.5: Output the fused dataset to the leak detection and localization module.

5. The wet membrane dialyzer membrane filament breakage detection system according to claim 1, characterized in that, The steps for analyzing the fused dataset and calculating the spatial location of the leak point based on the analysis results include: S3.1: Based on the fused dataset, calculate the data energy of all sensor channels and compare it with the leakage judgment threshold. When the data energy is greater than the leakage judgment threshold, generate a leakage signal and proceed to step S3.

2. S3.2: When a leakage signal is detected, the leakage trigger timestamp data is analyzed using a signal strength-based weighted centroid localization algorithm to obtain the leakage point coordinate data; S3.3: Pack the sensor ID of the leaking sensor, the coordinates of the leak point, the maximum data energy, the leak trigger timestamp, and the preset leak time window to obtain a leak event report; S3.4: Output the leakage event report to the membrane rupture feature extraction module.

6. The wet membrane dialyzer membrane filament breakage detection system according to claim 1, characterized in that, The steps for quantitative feature extraction based on leak event reports include: S4.1: Based on the preset leakage time window in the leakage event report, the feature vectors in the dataset corresponding to the time points of the preset leakage time window are fused in advance to obtain the truncated dataset; S4.2: Perform statistical calculations on the time-domain features in the extracted dataset to obtain a time-domain feature vector. The time-domain feature vector includes peak rise time feature, peak duration feature, and data distribution skewness feature, where: The peak rise time characteristic is obtained by statistically extracting the time required for the peak to rise from 10% to 90% in the dataset; The peak duration feature is obtained by statistically extracting the duration of more than 50% of the peaks in the dataset; The skewness characteristic of the data distribution is determined by substituting it into the calculation formula: We obtained, among which, To extract the number of feature vectors in the dataset, For the first 1 eigenvector To extract the mean of the feature vectors in the dataset, To extract the standard deviation of the feature vectors in the dataset; S4.3: Utilize the Fast Fourier Transform to calculate the frequency domain features in the truncated dataset, obtaining a frequency domain feature vector. This vector includes centroid frequency features and bandwidth features, where: The centroid frequency characteristic is the centroid frequency of the spectrum; The bandwidth feature is the bandwidth of the feature vector; S4.4: Combine the time-domain feature vector and the frequency-domain feature vector with a preset fixed length to obtain the breaking point feature vector; S4.5: Output the feature vector of the break point to the intelligent diagnosis module for the break point type.

7. The wet membrane dialyzer membrane filament breakage detection system according to claim 1, characterized in that, The steps for automatically identifying the fundamental type of a breakpoint by inputting the feature vector of the breakpoint into a pre-trained machine learning model include: S5.1: Load the pre-trained machine learning model; S5.2: Input the feature vector of the break point into the machine learning model, and output the predicted break point type and the confidence level of the prediction result; S5.3: Obtain a break point type diagnostic report by taking the packaging production equipment ID, predicted break point type, prediction result confidence level, and break point feature vector; S5.4: Output the failure point type diagnostic report to the root cause association analysis module.

8. The wet membrane dialyzer membrane filament breakage detection system according to claim 1, characterized in that, The steps for locating the cause in the production process based on the fault type diagnostic report and production batch data include: S6.1: Based on the database, query the relevant information of the current production batch to obtain the data for generating the current production batch. The relevant information includes material batch information, production equipment ID information, breakage type and operator ID information. S6.2: Based on the predicted break point type in the break point type diagnosis report, query the database for historical production batch data that matches the break point type to obtain the historical production batch dataset; S6.3: Match the predicted breach type in the breach type diagnosis report with the current production batch data, and perform root cause matching based on the historical production batch dataset in step S6.2 to obtain the leak root cause report; S6.4: Package the root cause report of the leak, the overlap, and the production parameter logs containing the production equipment ID information in the root cause report of the leak to obtain the root cause analysis report; S6.5: Output the root cause analysis report to the production parameter adaptive instruction generation module.

9. A wet membrane dialyzer membrane filament breakage detection system according to claim 2, characterized in that, The steps for a comprehensive diagnosis based on a root cause analysis report include: S7.1: Match the fault type diagnosis report and root cause analysis report with the preset adjustment strategy to obtain production parameter adjustment instructions; S7.2: Convert production parameter adjustment instructions into execution instruction format and output them to the intelligent controller; S7.3: Record the instruction issuance status and instruction execution feedback in step S7.2 to obtain an adjustment operation report; S7.4: Store the adjustment operation report in the database.

10. A method for detecting membrane filament breakage points in a wet membrane dialyzer, characterized in that, The wet membrane dialyzer membrane filament breakage detection system according to any one of claims 1-9 achieves the following working steps: S1: Perform unified scheduling of the start-up, shutdown, and parameters of the sensor network, and collect raw data to obtain the raw metadata dataset; S2: Preprocess and fuse the original metadata to obtain the fused dataset; S3: Analyze the fused dataset and calculate the spatial location of the leak point based on the analysis results to obtain a leak event report; S4: Quantitative feature extraction is performed based on the leak event report to obtain the feature vector of the breach point; S5: Input the feature vector of the breach point into the pre-trained machine learning model to automatically identify the fundamental type of the breach point and obtain a breach point type diagnosis report; S6: Based on the failure point type diagnostic report and production batch data, locate the cause in the production process and obtain the root cause analysis report; S7: Conduct a comprehensive diagnosis based on the root cause analysis report and obtain instructions for adjusting production parameters.

Images