A hemodialysis machine pump state early warning method and system based on audio recognition

Abstract

The application discloses a hemodialysis machine pump state early warning method and system based on audio recognition, and relates to the technical field of hemodialysis machine pump maintenance. The method comprises the following steps: collecting an audio signal of a hemodialysis machine pump; drawing a waveform graph based on the audio signal; preliminarily judging whether an abnormal waveform section is contained in the waveform graph based on image recognition; if the abnormal waveform section is contained in the waveform graph, obtaining an abnormal frequency spectrum graph of the abnormal waveform section; judging whether the hemodialysis machine pump is abnormal in combination with the abnormal waveform section and the abnormal frequency spectrum graph; and if the hemodialysis machine pump is abnormal, issuing an abnormal alarm. The application has the effect of detecting the state of the hemodialysis machine pump without disassembling the hemodialysis machine.

Description

Technical Field

[0001] This application relates to the field of hemodialysis pump maintenance technology, and in particular to a method and system for early warning of hemodialysis pump status based on audio recognition. Background Technology

[0002] Chronic renal failure refers to chronic, progressive damage to the renal parenchyma caused by various factors, leading to significant kidney atrophy and inability to maintain its basic functions. Chronic renal failure has become one of the most common chronic diseases. Clinically, the best treatment for chronic renal failure is kidney transplantation or dialysis. However, kidney transplantation not only requires high surgical costs but also necessitates matching with a suitable kidney, making it extremely difficult. Therefore, the vast majority of patients rely on hemodialysis machines for long-term dialysis to sustain their lives.

[0003] Because patients with chronic renal failure require frequent dialysis sessions, and each session takes a long time, overuse of the hemodialysis machine can lead to severe wear and tear on the hemodialysis pump. If the pump malfunctions during dialysis, it can endanger the patient's life. Therefore, in related technologies, it is usually necessary to manually inspect the hemodialysis pump regularly to reduce the impact of severe wear and tear or aging on the operation of the hemodialysis machine.

[0004] Regarding the aforementioned technologies, the inventors believe that the following drawbacks exist: manual maintenance of the hemodialysis pump requires disassembling the hemodialysis machine, and frequent disassembly of the hemodialysis machine can easily increase the wear and tear on the hemodialysis machine. Summary of the Invention

[0005] To address the drawback of manual maintenance of hemodialysis pumps, which can easily increase the wear and tear on the hemodialysis machine, this application provides a hemodialysis pump status early warning method and system based on audio recognition.

[0006] In a first aspect, this application provides a method for early warning of the status of a hemodialysis pump based on audio recognition, comprising the following steps:

[0007] Collect audio signals from the hemodialysis machine pump during operation;

[0008] Draw a waveform diagram based on the audio signal;

[0009] Based on image recognition, a preliminary determination is made as to whether the waveform image contains abnormal waveform segments;

[0010] If the waveform diagram contains the abnormal waveform segment, then obtain the abnormal spectrum diagram of the abnormal waveform segment;

[0011] By combining the abnormal waveform segments and the abnormal spectrum diagram, it can be determined whether the hemodialysis pump is malfunctioning;

[0012] If the hemodialysis pump malfunctions, an alarm will be issued.

[0013] By employing the above technical solution, the audio signals from the hemodialysis pump during operation are collected and waveforms are plotted. Image recognition technology can be used to initially determine whether abnormal waveform segments are present in the waveform diagram. If abnormal waveform segments are found, further analysis can be conducted by combining the abnormal waveform segments with the abnormal spectrum diagram to determine if there is an abnormality in the hemodialysis pump. If an abnormality is detected, an alarm is triggered to remind staff to promptly inspect and maintain the hemodialysis pump. Compared to manual periodic maintenance in related technologies, the operating status of the hemodialysis pump can be analyzed and judged solely based on its operating sound, without the need for frequent disassembly of the hemodialysis machine. Therefore, this can reduce the occurrence of hemodialysis pump failures and extend the overall service life of the hemodialysis machine.

[0014] Optionally, the preliminary determination of whether the waveform image contains abnormal waveform segments based on image recognition includes the following steps:

[0015] A basic recognition model is built based on the YOLOv3 model;

[0016] Retrieve multiple historical abnormal waveform segments from the preset waveform database;

[0017] The historical abnormal waveform segments are preprocessed to obtain preprocessed waveform segments;

[0018] The basic recognition model is trained using the preprocessed waveform segments to obtain the anomaly recognition model;

[0019] The waveform is imported into the anomaly identification model, and the anomaly identification model is used to initially determine whether the waveform contains abnormal waveform segments.

[0020] By adopting the above technical solution, after constructing a basic waveform recognition model based on the YOLOv3 model, historical abnormal waveform segments can be retrieved from the waveform database to train the basic recognition model, so as to train the optimal abnormal recognition model. Then, the waveform diagram can be analyzed and judged through the abnormal recognition model, so as to preliminarily determine whether the waveform diagram contains abnormal waveform segments that are suspected of abnormal sounds.

[0021] Optionally, determining whether the hemodialysis pump is malfunctioning by combining the abnormal waveform segment and the abnormal spectrum diagram includes the following steps:

[0022] Determine whether there are any abrupt changes in the abnormal waveform segment;

[0023] If the abrupt change point is not present in the abnormal waveform segment, then the abnormal spectrum diagram is used to determine whether the hemodialysis pump is abnormal.

[0024] If the abrupt change point exists in the abnormal waveform segment, then the kurtosis value of the abnormal waveform segment is calculated;

[0025] Determine whether the kurtosis value is lower than a preset kurtosis threshold;

[0026] If the kurtosis value is lower than the kurtosis threshold, then the abnormal spectrum diagram is used to determine whether the hemodialysis pump is abnormal;

[0027] If the kurtosis value is higher than the kurtosis threshold, the hemodialysis pump is determined to be malfunctioning.

[0028] By adopting the above technical solution, it is possible to determine whether there is instantaneous noise during the operation of the hemodialysis pump by judging whether there is abrupt change in the abnormal waveform segment. If there is abrupt change, it indicates that the hemodialysis pump may have abnormalities such as loose parts. It is necessary to further calculate the kurtosis value of the abnormal waveform segment, and then determine whether the hemodialysis pump has abnormalities by comparing the preset kurtosis threshold with the kurtosis value. If there is no abrupt change in the abnormal waveform segment or the kurtosis value is lower than the kurtosis threshold, it is necessary to further analyze and judge the hemodialysis pump based on the abnormal spectrum diagram.

[0029] Optionally, the formula for calculating the kurtosis value of the abnormal waveform segment is:

[0030]

[0031] In the formula, x i Let be the i-th signal value in the abnormal waveform segment, x be the signal mean of the abnormal waveform segment, and σ be the standard deviation of all signals in the abnormal waveform segment.

[0032] By adopting the above technical solution, when the hemodialysis pump is in normal operation, the kurtosis value calculated by the above formula will be stable in a constant range. If the hemodialysis pump experiences pulse impact caused by component damage or aging during operation, the kurtosis value will increase.

[0033] Optionally, determining whether the hemodialysis pump is abnormal based on the abnormal spectrum diagram includes the following steps:

[0034] Retrieve multiple historical normal waveform segments from the waveform database;

[0035] Adjust the signal length of all historical normal waveform segments based on the abnormal waveform segment;

[0036] Obtain the normal spectrum diagrams of all the historical normal waveform segments after adjustment, wherein the normal spectrum diagrams and the abnormal spectrum diagrams have the same spectral resolution;

[0037] The amplitude is used as a feature value to enclose the abnormal spectrum and all the normal spectrum, resulting in an abnormal envelope and multiple normal envelopes;

[0038] A reference frequency range is defined based on multiple normal envelope maps, and the average area of ​​the multiple normal envelope maps in the reference frequency range is calculated.

[0039] Calculate the area of ​​the special interval in the reference frequency range of the abnormal envelope diagram;

[0040] Determine whether the area of ​​the special interval is less than the average interval area;

[0041] If the area of ​​the special interval is smaller than the area of ​​the average interval, then the hemodialysis pump is determined to be abnormal.

[0042] By adopting the above technical solution, when using the spectrum diagram for anomaly analysis, the normal spectrum diagram can be comprehensively analyzed first to define the reference frequency range and calculate the average area of ​​the range between the envelope and the horizontal axis. Then, the area of ​​the special range within the same range in the abnormal envelope diagram can be calculated. By comparing the size of the special range area and the average range area, it can be determined whether there is an abnormality in the hemodialysis pump.

[0043] Optionally, after determining that the hemodialysis pump is malfunctioning, the following steps are also included:

[0044] Label the abnormal waveform segments with an abnormal tag;

[0045] The abnormal waveform segments in the waveform diagram are filtered out, and the normal waveforms in the waveform diagram are retained;

[0046] The normal waveform is divided into multiple normal waveform segments according to a preset standard signal length;

[0047] Each of the aforementioned normal waveform segments is labeled with a normal tag;

[0048] The abnormal waveform segments and all the normal waveform segments are stored in the waveform database.

[0049] By adopting the above technical solution, if the hemodialysis pump is found to be abnormal, the collected abnormal waveform segments can be stored as abnormal samples in the waveform database, which is beneficial for the subsequent training of the abnormal identification model, thereby improving the accuracy of the abnormal identification model. Normal waveforms can also be stored as normal samples in the waveform database, which is beneficial for improving the identification accuracy of subsequent spectrum anomaly analysis.

[0050] Optionally, obtaining the abnormal spectrum of the abnormal waveform segment includes the following steps:

[0051] The abnormal waveform segment is processed by ADC to obtain abnormal digital samples;

[0052] The abnormal digital samples are subjected to a Fast Fourier Transform based on a preset sampling frequency to obtain an abnormal spectrum.

[0053] By adopting the above technical solution, abnormal waveform segments can be processed by ADC before performing Fast Fourier Transform, which helps to improve the accuracy of Fast Fourier Transform.

[0054] Secondly, this application also provides an audio recognition-based hemodialysis pump status early warning system, including a memory, a processor, and a program stored in the memory and executable on the processor. When the program is loaded and executed by the processor, it can implement an audio recognition-based hemodialysis pump status early warning method as described in the first aspect.

[0055] By employing the above technical solution, the audio signal of the hemodialysis pump during operation can be collected and a waveform diagram can be generated through program retrieval. Image recognition technology can be used to initially determine whether the waveform diagram contains abnormal waveform segments. If abnormal waveform segments are found, further analysis can be performed by combining the abnormal waveform segments with the abnormal spectrum diagram to determine if the hemodialysis pump is malfunctioning. If an abnormality is detected, an alarm is issued to remind staff to promptly inspect and maintain the hemodialysis pump. Compared to manual periodic maintenance in related technologies, the operating status of the hemodialysis pump can be analyzed and judged solely based on its operating sound, without the need for frequent disassembly of the hemodialysis machine. Therefore, this can reduce the occurrence of hemodialysis pump failures and extend the overall service life of the hemodialysis machine.

[0056] In summary, this application includes the following beneficial technical effects:

[0057] By collecting the audio signals from the hemodialysis pump during operation and plotting the waveform, image recognition technology can be used to initially determine whether abnormal waveform segments are present. If abnormal waveform segments are found, further analysis combining the abnormal waveform segments and abnormal spectrum diagrams can be performed to determine if there is an abnormality in the hemodialysis pump. If an abnormality is detected, an alarm is triggered to remind staff to promptly inspect and maintain the hemodialysis pump. Compared to manual periodic maintenance in related technologies, the operating status of the hemodialysis pump can be analyzed and judged solely based on its operating sound, without the need for frequent disassembly of the hemodialysis machine. Therefore, this can reduce the occurrence of hemodialysis pump failures and extend the overall lifespan of the hemodialysis machine. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating one embodiment of the audio recognition-based hemodialysis pump status early warning method according to this application.

[0059] Figure 2 This is a flowchart illustrating one embodiment of the audio recognition-based hemodialysis pump status early warning method according to this application.

[0060] Figure 3 This is a flowchart illustrating one embodiment of the audio recognition-based hemodialysis pump status early warning method according to this application.

[0061] Figure 4 This is a flowchart illustrating one embodiment of the audio recognition-based hemodialysis pump status early warning method according to this application.

[0062] Figure 5 This is a flowchart illustrating one embodiment of the audio recognition-based hemodialysis pump status early warning method according to this application. Detailed Implementation

[0063] The following is in conjunction with the appendix Figures 1 to 5 This application will be described in further detail.

[0064] This application discloses a method for early warning of the status of a hemodialysis machine pump based on audio recognition.

[0065] Reference Figure 1 The audio recognition-based early warning method for the status of hemodialysis pumps includes the following steps:

[0066] S101. Collect audio signals during the operation of the hemodialysis pump.

[0067] Among them, the audio signal of the hemodialysis pump during operation can be collected by a microphone that is pre-set next to the hemodialysis pump.

[0068] S102. Draw a waveform diagram based on the audio signal.

[0069] Before drawing the waveform, the audio signal needs to be preprocessed. The preprocessing steps include extracting the audio signal based on a preset duration, then denoising the extracted audio signal, and finally drawing the waveform based on the denoised audio signal. The waveform is a time sequence diagram, with time on the horizontal axis and amplitude on the vertical axis.

[0070] S103. Based on image recognition, preliminarily determine whether the waveform diagram contains abnormal waveform segments. If it contains abnormal waveform segments, proceed to step S104.

[0071] When the hemodialysis pump is in normal operation, the waveform is relatively stable with the amplitude below 0.6V. However, when the hemodialysis pump malfunctions or other abnormal phenomena, the amplitude of the waveform will change abruptly, and there will be abnormal waveform segments with an overall average amplitude exceeding 0.6V. Since the abnormal waveform segments and the normal waveform segments are quite different in the graph, image recognition can be used to make a preliminary identification and judgment of the abnormal waveform segments in the waveform graph.

[0072] S104. Obtain the abnormal spectrum diagram of the abnormal waveform segment.

[0073] The specific steps for obtaining the abnormal spectrum diagram of the abnormal waveform segment are as follows:

[0074] S104A. Perform ADC processing on the abnormal waveform segment to obtain abnormal digital samples;

[0075] ACD processing is a type of filtering and sampling processing.

[0076] S104B. Perform a Fast Fourier Transform on the abnormal digital samples based on a preset sampling frequency to obtain an abnormal spectrum.

[0077] The time domain can be converted to the frequency domain by using the Fast Fourier Transform, thus obtaining the abnormal spectrum diagram. The horizontal axis of the spectrum diagram is frequency, and the vertical axis is amplitude.

[0078] S105. Determine whether there is an abnormality in the hemodialysis pump by combining the abnormal waveform segment and the abnormal spectrum diagram. If there is an abnormality in the hemodialysis pump, proceed to step S106.

[0079] S106. Issue an abnormal alarm.

[0080] The abnormal alarms include triggering the buzzer installed in the hemodialysis machine to sound an alarm, and also sending alarm information to the mobile terminals held by relevant staff.

[0081] The implementation principle of this embodiment is as follows:

[0082] By collecting the audio signals from the hemodialysis pump during operation and plotting the waveform, image recognition technology can be used to initially determine whether abnormal waveform segments are present. If abnormal waveform segments are found, further analysis combining the abnormal waveform segments and abnormal spectrum diagrams can be performed to determine if there is an abnormality in the hemodialysis pump. If an abnormality is detected, an alarm is triggered to remind staff to promptly inspect and maintain the hemodialysis pump. Compared to manual periodic maintenance in related technologies, the operating status of the hemodialysis pump can be analyzed and judged solely based on its operating sound, without the need for frequent disassembly of the hemodialysis machine. Therefore, this can reduce the occurrence of hemodialysis pump failures and extend the overall lifespan of the hemodialysis machine.

[0083] exist Figure 1 In step S103 of the illustrated embodiment, a basic recognition model can be built based on YOLOv3, and then the model can be trained into an optimal model using historical abnormal waveform segments. Finally, the optimal model is used to perform preliminary recognition of the waveform. Specifically, through... Figure 2 The illustrated embodiments will be described in detail.

[0084] Reference Figure 2The preliminary determination of whether an abnormal waveform segment is contained in a waveform image based on image recognition includes the following steps:

[0085] S201. Construct a basic recognition model based on the YOLOv3 model.

[0086] Among them, the YOLOv3 model is a target detection model, which can be used to build a basic recognition model for identifying abnormal waveform segments.

[0087] S202. Retrieve multiple historical abnormal waveform segments from the preset waveform database.

[0088] The waveform database pre-stores a large number of historical waveform segments. All historical waveform segments have the same signal length and include historical abnormal waveform segments and historical normal waveform segments. The historical abnormal waveform segments are the waveforms of the hemodialysis pump when it was in an abnormal operating state, while the historical normal waveform segments are the waveforms of the hemodialysis pump when it was in a normal operating state. The historical waveform segments in the waveform database are derived from waveforms processed from audio signals historically collected by microphones, as well as waveforms processed from operating audio signals of the same model of hemodialysis pump retrieved from the Internet.

[0089] S203. Preprocess historical abnormal waveform segments to obtain preprocessed waveform segments.

[0090] The main task is to preprocess the waveform images of historical abnormal waveform segments. The preprocessing steps include noise reduction and image enhancement.

[0091] S204. Train the basic recognition model by preprocessing waveform segments to obtain the anomaly recognition model.

[0092] In this process, a large number of preprocessed waveform segments are imported into the basic recognition model as training examples. Through model training, each parameter in the basic recognition model is trained to the optimal parameter. The basic recognition model after training is the anomaly recognition model.

[0093] S205. Import the waveform diagram into the anomaly identification model, and use the anomaly identification model to initially determine whether the waveform diagram contains abnormal waveform segments.

[0094] Among them, the amplitude characteristics of the waveform are identified by the anomaly identification model, and waveform segments with abrupt amplitude changes and waveform segments with an overall average amplitude exceeding a certain threshold are identified as abnormal waveform segments.

[0095] The implementation principle of this embodiment is as follows:

[0096] After building a basic waveform recognition model based on the YOLOv3 model, historical abnormal waveform segments can be retrieved from the waveform database to train the basic recognition model, so as to train the optimal abnormal recognition model. Then, the waveform diagram can be analyzed and judged through the abnormal recognition model, so as to preliminarily determine whether the waveform diagram contains abnormal waveform segments that are suspected of abnormal sounds.

[0097] exist Figure 1 In step S105 of the illustrated embodiment, a judgment is first made based on the waveform characteristics and kurtosis value of the abnormal waveform segment. If no abnormality is determined in the hemodialysis pump based on the abnormal waveform segment, further analysis and judgment are performed based on the abnormal spectrum diagram. Specifically, through... Figure 3 The illustrated embodiments will be described in detail.

[0098] Reference Figure 3 Determining whether a hemodialysis pump is malfunctioning by combining abnormal waveform segments and abnormal spectrum diagrams includes the following steps:

[0099] S301. Determine whether there is a sudden change point in the abnormal waveform segment. If there is no sudden change point in the abnormal waveform segment, proceed to step S302; if there is a sudden change point in the abnormal waveform segment, proceed to step S303.

[0100] The system uses a preset mutation threshold to determine if the amplitude at any node in the abnormal waveform segment exceeds the preset mutation threshold. This indicates that there is a mutation point in the abnormal waveform segment.

[0101] S302. Determine if there is any abnormality in the hemodialysis pump based on the abnormal spectrum diagram.

[0102] S303. Calculate the kurtosis value of the abnormal waveform segment.

[0103] The formula for calculating the kurtosis value of abnormal waveform segments is as follows:

[0104]

[0105] In the formula, x i Let denoted as the i-th signal value in the abnormal waveform segment, x be the signal mean of the abnormal waveform segment, and σ be the standard deviation of all signals in the abnormal waveform segment. When the hemodialysis pump is operating normally, the kurtosis value calculated using the above formula will stabilize within a constant range. However, if the hemodialysis pump experiences pulse impacts due to component damage or aging during operation, the kurtosis value will increase.

[0106] S304. Determine whether the kurtosis value is lower than the preset kurtosis threshold. If the kurtosis value is lower than the kurtosis threshold, proceed to step S305; if the kurtosis value is higher than the kurtosis threshold, proceed to step S306.

[0107] In the time domain, kurtosis is a commonly used dimensionless characteristic parameter for abnormal signal analysis. It is sensitive to pulse impact components, and damage or aging of equipment components can be determined by analyzing kurtosis values. During normal operation, the kurtosis value of a hemodialysis pump is typically maintained at around 3, while it rises to above 4 when the pump is in an abnormal operating state. Therefore, the kurtosis threshold is usually preset to 4. If the kurtosis value equals the kurtosis threshold, step S305 is executed.

[0108] S305. Determine if there is an abnormality in the hemodialysis pump based on the abnormal spectrum diagram.

[0109] S306. The hemodialysis machine pump is found to be malfunctioning.

[0110] The implementation principle of this embodiment is as follows:

[0111] By judging whether there are abrupt changes in the abnormal waveform segment, it is determined whether there is instantaneous noise during the operation of the hemodialysis pump. If there are abrupt changes, it indicates that the hemodialysis pump may have abnormalities such as loose parts. It is necessary to further calculate the kurtosis value of the abnormal waveform segment, and then compare the preset kurtosis threshold with the kurtosis value to determine whether the hemodialysis pump has abnormalities. If there are no abrupt changes in the abnormal waveform segment or the kurtosis value is lower than the kurtosis threshold, it is necessary to further analyze and judge the hemodialysis pump based on the abnormal spectrum diagram.

[0112] exist Figure 3 In step S302 or step S305 of the illustrated embodiment, a normal envelope diagram and an abnormal envelope diagram are first drawn based on the normal and abnormal spectrum diagrams. A reference frequency range is defined based on the normal spectrum diagram, and the average range area is calculated. Then, the area of ​​a specific range in the abnormal spectrum diagram is calculated. By comparing the area sizes, it is possible to determine whether the hemodialysis pump is malfunctioning. Specifically, through... Figure 4 The illustrated embodiments will be described in detail.

[0113] Reference Figure 4 Determining whether a hemodialysis pump is malfunctioning based on an abnormal spectrum diagram involves the following steps:

[0114] S401. Retrieve multiple historical normal waveform segments from the waveform database.

[0115] In this context, all historical normal waveform segments in the waveform database are marked with a normal label. Therefore, multiple historical normal waveform segments can be retrieved from the waveform database based on the normal label, and these multiple historical normal waveform segments can be retrieved.

[0116] S402. Adjust the signal length of all historical normal waveform segments based on the abnormal waveform segments.

[0117] In order to facilitate subsequent frequency domain analysis, it is necessary to first unify the signal lengths of abnormal waveform segments and historical normal waveform segments.

[0118] S403. Obtain the normal spectrum diagram of all historical normal waveform segments after adjustment.

[0119] The normal spectrum can be obtained by referring to steps S104A to S104B. The normal spectrum and the abnormal spectrum have the same spectral resolution.

[0120] S404. Using amplitude as the eigenvalue, encapsulate the abnormal spectrum and all normal spectrums to obtain the abnormal envelope and multiple normal envelopes.

[0121] In this process, N consecutive points in the abnormal spectrum are divided into a group, and M groups are formed in the entire abnormal spectrum. The maximum value of the amplitude of N points in each group is taken as the characteristic value of that group. The characteristic values ​​of the M groups are then connected sequentially with curves to obtain the abnormal envelope of the entire abnormal spectrum. The normal spectrum is obtained in the same way.

[0122] S405. Delineate a reference frequency range based on multiple normal envelope diagrams, and calculate the average area of ​​the multiple normal envelope diagrams within the reference frequency range.

[0123] By analyzing multiple normal envelope diagrams, a reference frequency range can be defined. Within this range, the amplitude fluctuations in normal envelope diagrams are relatively large, while the amplitude fluctuations in abnormal envelope diagrams are relatively small. Calculating the average area of ​​multiple normal envelope diagrams within the reference frequency range involves first calculating the basic area of ​​each normal envelope diagram. The basic area is the area enclosed by the horizontal axis of the envelope diagram, the envelope line within the reference frequency range, and the two interval lines of the reference frequency range. Assuming the reference frequency range is [1000 Hz, 2000 Hz], then the two interval lines are x = 1000 and x = 2000. After calculating the basic area of ​​each normal envelope diagram, the average of the basic area of ​​all normal envelope diagrams is calculated; this average area is the total area.

[0124] S406. Calculate the area of ​​the special interval of the abnormal envelope in the reference frequency range.

[0125] The calculation method for the area of ​​special intervals is detailed in step S405.

[0126] S407. Determine whether the area of ​​the special interval is less than the area of ​​the average interval. If the area of ​​the special interval is less than the area of ​​the average interval, proceed to step S408.

[0127] If the area of ​​a special interval is greater than or equal to the area of ​​the average interval, the hemodialysis pump is considered to be functioning normally.

[0128] S408. The hemodialysis machine pump is found to be malfunctioning.

[0129] The implementation principle of this embodiment is as follows:

[0130] When using a spectrum diagram for anomaly analysis, a comprehensive analysis of the normal spectrum diagram can be performed first to define the reference frequency range and calculate the average area of ​​the interval between the envelope and the horizontal axis. Then, the area of ​​the special interval within the same interval in the abnormal envelope diagram can be calculated. By comparing the size of the special interval area and the average interval area, it can be determined whether there is an abnormality in the hemodialysis pump.

[0131] exist Figure 3 Step S306 of the illustrated embodiment or Figure 4 In step S408 of the illustrated embodiment, after determining that the hemodialysis pump is malfunctioning, the acquired waveform can be separated into abnormal waveform segments and normal waveform segments, and stored separately in the waveform database. Specifically, through... Figure 5 The illustrated embodiments will be described in detail.

[0132] Reference Figure 5 After determining that the hemodialysis pump is malfunctioning, the following steps are also included:

[0133] S501. Label abnormal waveform segments with an abnormal label.

[0134] S502. Filter out abnormal waveform segments in the waveform diagram and retain the normal waveforms in the waveform diagram.

[0135] In this process, abnormal waveform segments are removed from the waveform diagram, and the remaining normal waveforms are spliced ​​together and preserved.

[0136] S503. Divide the normal waveform into multiple normal waveform segments according to the preset standard signal length.

[0137] S504. Label multiple normal waveform segments with normal labels respectively.

[0138] S505. Store abnormal waveform segments and all normal waveform segments in the waveform database.

[0139] The implementation principle of this embodiment is as follows:

[0140] If the hemodialysis pump is found to be malfunctioning, the collected abnormal waveform segments can be stored as abnormal samples in the waveform database. This is beneficial for training the anomaly identification model and thus improving the accuracy of the anomaly identification model. Normal waveforms can also be stored as normal samples in the waveform database, which is beneficial for improving the identification accuracy of subsequent spectrum anomaly analysis.

[0141] This application also discloses an audio recognition-based hemodialysis pump status early warning system, including a memory, a processor, and a program stored in the memory and executable on the processor. This program, when loaded and executed by the processor, can achieve the following: Figures 1 to 5The image shows a method for early warning of the status of a hemodialysis machine pump based on audio recognition.

[0142] The implementation principle of this embodiment is as follows:

[0143] By retrieving the program, the audio signal of the hemodialysis pump during operation can be collected and a waveform diagram can be generated. Image recognition technology can be used to initially determine if abnormal waveform segments are present. If abnormal waveform segments are found, further analysis combining the abnormal waveform segments and abnormal spectrum diagrams can be performed to determine if the hemodialysis pump is malfunctioning. If an abnormality is detected, an alarm is issued to remind staff to promptly inspect and maintain the hemodialysis pump. Compared to manual periodic maintenance in related technologies, the operating status of the hemodialysis pump can be analyzed and judged solely based on its operating sound, without the need for frequent disassembly of the hemodialysis machine. Therefore, this can reduce the occurrence of hemodialysis pump failures and extend the overall lifespan of the hemodialysis machine.

[0144] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for early warning of hemodialysis machine pump status based on audio recognition, characterized in that, Includes the following steps: Collect audio signals from the hemodialysis machine pump during operation; Draw a waveform diagram based on the audio signal; Based on image recognition, a preliminary determination is made as to whether the waveform image contains abnormal waveform segments; If the waveform diagram contains the abnormal waveform segment, then obtain the abnormal spectrum diagram of the abnormal waveform segment; By combining the abnormal waveform segments and the abnormal spectrum diagram, it can be determined whether the hemodialysis pump is malfunctioning; If the hemodialysis pump malfunctions, an alarm will be issued. The preliminary determination of whether the waveform image contains abnormal waveform segments based on image recognition includes the following steps: A basic recognition model is built based on the YOLOv3 model; Retrieve multiple historical abnormal waveform segments from the preset waveform database; The historical abnormal waveform segments are preprocessed to obtain preprocessed waveform segments; The basic recognition model is trained using the preprocessed waveform segments to obtain the anomaly recognition model; The waveform is imported into the anomaly detection model, and the anomaly detection model is used to initially determine whether the waveform contains abnormal waveform segments. The step of determining whether the hemodialysis pump is malfunctioning by combining the abnormal waveform segment and the abnormal spectrum diagram includes the following steps: Determine whether there are any abrupt changes in the abnormal waveform segment; If the abrupt change point is not present in the abnormal waveform segment, then the abnormal spectrum diagram is used to determine whether the hemodialysis pump is abnormal. If the abrupt change point exists in the abnormal waveform segment, then the kurtosis value of the abnormal waveform segment is calculated; Determine whether the kurtosis value is lower than a preset kurtosis threshold; If the kurtosis value is lower than the kurtosis threshold, then the abnormal spectrum diagram is used to determine whether the hemodialysis pump is abnormal; If the kurtosis value is higher than the kurtosis threshold, the hemodialysis pump is determined to be malfunctioning. The step of determining whether the hemodialysis pump is abnormal based on the abnormal spectrum diagram includes the following steps: Retrieve multiple historical normal waveform segments from the waveform database; Adjust the signal length of all historical normal waveform segments based on the abnormal waveform segment; Obtain the normal spectrum diagrams of all the historical normal waveform segments after adjustment, wherein the normal spectrum diagrams and the abnormal spectrum diagrams have the same spectral resolution; The amplitude is used as a feature value to enclose the abnormal spectrum and all the normal spectrum, resulting in an abnormal envelope and multiple normal envelopes; A reference frequency range is defined based on multiple normal envelope maps, and the average area of ​​the multiple normal envelope maps in the reference frequency range is calculated. Calculate the area of ​​the special interval in the reference frequency range of the abnormal envelope diagram; Determine whether the area of ​​the special interval is less than the average interval area; If the area of ​​the special interval is smaller than the area of ​​the average interval, then the hemodialysis pump is determined to be abnormal.

2. The method for early warning of hemodialysis machine pump status based on audio recognition according to claim 1, characterized in that: The formula for calculating the kurtosis value of the abnormal waveform segment is as follows: In the formula, x i Let i be the i-th signal value in the abnormal waveform segment. Let σ be the signal mean of the abnormal waveform segment, and σ be the standard deviation of all signals in the abnormal waveform segment.

3. The method for early warning of hemodialysis machine pump status based on audio recognition according to claim 1, characterized in that, After determining that the hemodialysis pump is malfunctioning, the following steps are also included: Label the abnormal waveform segments with an abnormal tag; The abnormal waveform segments in the waveform diagram are filtered out, and the normal waveforms in the waveform diagram are retained; The normal waveform is divided into multiple normal waveform segments according to a preset standard signal length; Each of the aforementioned normal waveform segments is labeled with a normal tag; The abnormal waveform segments and all the normal waveform segments are stored in the waveform database.

4. The method for early warning of hemodialysis machine pump status based on audio recognition according to claim 1, characterized in that, The steps for obtaining the abnormal spectrum diagram of the abnormal waveform segment include the following: The abnormal waveform segment is processed by ADC to obtain abnormal digital samples; The abnormal digital samples are subjected to a Fast Fourier Transform based on a preset sampling frequency to obtain an abnormal spectrum.

5. A hemodialysis machine pump status early warning system based on audio recognition, characterized in that, It includes a memory, a processor, and a program stored in the memory and executable on the processor, which, when loaded and executed by the processor, implements an audio recognition-based hemodialysis pump status early warning method as described in any one of claims 1 to 4.

Images