CN115471827A - Hemodialysis machine pump state early warning method and system based on audio recognition - Google Patents
Hemodialysis machine pump state early warning method and system based on audio recognition Download PDFInfo
- Publication number
- CN115471827A CN115471827A CN202211034978.0A CN202211034978A CN115471827A CN 115471827 A CN115471827 A CN 115471827A CN 202211034978 A CN202211034978 A CN 202211034978A CN 115471827 A CN115471827 A CN 115471827A
- Authority
- CN
- China
- Prior art keywords
- abnormal
- waveform
- pump
- normal
- spectrogram
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001631 haemodialysis Methods 0.000 title claims abstract description 106
- 230000000322 hemodialysis Effects 0.000 title claims abstract description 106
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000002159 abnormal effect Effects 0.000 claims abstract description 251
- 230000005236 sound signal Effects 0.000 claims abstract description 20
- 239000008280 blood Substances 0.000 claims description 34
- 210000004369 blood Anatomy 0.000 claims description 34
- 238000010586 diagram Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 9
- 230000035772 mutation Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 4
- 238000001228 spectrum Methods 0.000 claims description 4
- 230000005856 abnormality Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000012423 maintenance Methods 0.000 abstract description 7
- 230000000694 effects Effects 0.000 abstract description 2
- 238000000502 dialysis Methods 0.000 description 6
- 230000002035 prolonged effect Effects 0.000 description 5
- 230000032683 aging Effects 0.000 description 4
- 208000020832 chronic kidney disease Diseases 0.000 description 4
- 208000022831 chronic renal failure syndrome Diseases 0.000 description 4
- 210000003734 kidney Anatomy 0.000 description 4
- 230000006378 damage Effects 0.000 description 2
- 238000002054 transplantation Methods 0.000 description 2
- 206010003694 Atrophy Diseases 0.000 description 1
- 208000017667 Chronic Disease Diseases 0.000 description 1
- 230000037444 atrophy Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000740 bleeding effect Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 230000001684 chronic effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
- A61M1/00—Suction or pumping devices for medical purposes; Devices for carrying-off, for treatment of, or for carrying-over, body-liquids; Drainage systems
- A61M1/14—Dialysis systems; Artificial kidneys; Blood oxygenators ; Reciprocating systems for treatment of body fluids, e.g. single needle systems for hemofiltration or pheresis
- A61M1/16—Dialysis systems; Artificial kidneys; Blood oxygenators ; Reciprocating systems for treatment of body fluids, e.g. single needle systems for hemofiltration or pheresis with membranes
- A61M1/1601—Control or regulation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
- A61M60/00—Blood pumps; Devices for mechanical circulatory actuation; Balloon pumps for circulatory assistance
- A61M60/10—Location thereof with respect to the patient's body
- A61M60/104—Extracorporeal pumps, i.e. the blood being pumped outside the patient's body
- A61M60/109—Extracorporeal pumps, i.e. the blood being pumped outside the patient's body incorporated within extracorporeal blood circuits or systems
- A61M60/113—Extracorporeal pumps, i.e. the blood being pumped outside the patient's body incorporated within extracorporeal blood circuits or systems in other functional devices, e.g. dialysers or heart-lung machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
- G06V10/422—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation for representing the structure of the pattern or shape of an object therefor
- G06V10/426—Graphical representations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
- G08B21/24—Reminder alarms, e.g. anti-loss alarms
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Heart & Thoracic Surgery (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Veterinary Medicine (AREA)
- Cardiology (AREA)
- Evolutionary Computation (AREA)
- Anesthesiology (AREA)
- Biomedical Technology (AREA)
- Hematology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Urology & Nephrology (AREA)
- Public Health (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mechanical Engineering (AREA)
- Business, Economics & Management (AREA)
- Emergency Medicine (AREA)
- Vascular Medicine (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Pulmonology (AREA)
- Emergency Management (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- External Artificial Organs (AREA)
Abstract
The application discloses hemodialysis machine pump state early warning method and system based on audio recognition, and relates to the technical field of hemodialysis machine pump maintenance, and the method comprises the following steps: collecting an audio signal of a hemodialysis machine pump; drawing a waveform map based on the audio signal; preliminarily judging whether the oscillogram contains an abnormal waveform section or not based on image identification; if the oscillogram contains the abnormal waveform section, acquiring an abnormal frequency spectrogram of the abnormal waveform section; judging whether the hemodialysis pump is abnormal or not by combining the abnormal waveform segment and the abnormal spectrogram; and if the hemodialysis machine pump is abnormal, an abnormal alarm is sent out. This application has the effect that need not dismantle hemodialysis machine and can detect hemodialysis machine pump state.
Description
Technical Field
The application relates to the technical field of hemodialysis machine pump maintenance, in particular to a hemodialysis machine pump state early warning method and system based on audio recognition.
Background
Chronic renal failure refers to chronic progressive damage to the renal parenchyma caused by various causes, resulting in significant atrophy of the kidney and failure to maintain its basic function. Chronic renal failure has become one of several common chronic diseases, the clinically best treatment method for chronic renal failure is kidney transplantation or dialysis, but kidney transplantation not only requires high operation cost, but also needs matching with transplantable kidneys, and is very difficult, so that most patients adopt a hemodialysis machine for long-term dialysis to maintain life.
Because the dialysis frequency required by a patient with chronic renal failure is high, and each dialysis needs a long time, the excessive use of the hemodialysis machine easily causes the serious loss of a hemodialysis pump in the dialysis machine, and if the hemodialysis pump fails in the dialysis process, the life safety of the patient is easily endangered. Therefore, in the related art, the hemodialysis machine pump usually needs to be manually and periodically overhauled, so as to reduce the influence on the operation of the hemodialysis machine caused by serious wear or aging of the hemodialysis machine pump during use.
With respect to the related art among the above, the inventors consider that the following drawbacks exist: the hemodialysis machine needs to be disassembled when the hemodialysis machine pump is manually overhauled, and the loss degree of the hemodialysis machine is easily increased when the hemodialysis machine is frequently disassembled.
Disclosure of Invention
In order to improve the defect that the hemodialysis machine pump is manually overhauled to easily increase the loss degree of the hemodialysis machine, the application provides a hemodialysis machine pump state early warning method and system based on audio identification.
In a first aspect, the application provides a hemodialysis machine pump state early warning method based on audio recognition, which includes the following steps:
collecting audio signals when the hemodialysis machine pump runs;
drawing a waveform map based on the audio signal;
preliminarily judging whether the oscillogram contains an abnormal waveform section or not based on image identification;
if the oscillogram contains the abnormal waveform section, acquiring an abnormal frequency spectrogram of the abnormal waveform section;
judging whether the hemodialysis pump is abnormal or not by combining the abnormal waveform segment and the abnormal spectrogram;
and if the hemodialysis machine pump is abnormal, an abnormal alarm is sent out.
By adopting the technical scheme, the audio signals generated when the blood turbine pump runs are collected and the oscillogram is drawn, whether the oscillogram contains an abnormal waveform section or not can be preliminarily judged according to the image recognition technology, if the oscillogram contains the abnormal waveform section, the abnormal waveform section and the abnormal frequency spectrogram can be combined for further analysis and judgment to judge whether the blood turbine pump is abnormal or not, and if the abnormal waveform section is detected out, an abnormal alarm is sent out to remind a worker to overhaul the blood turbine pump in time. Compared with manual regular maintenance in the related art, the running state of the hemodialysis machine pump can be analyzed and judged only by the running sound of the hemodialysis machine pump, and the hemodialysis machine does not need to be frequently disassembled, so that the faults of the hemodialysis machine pump can be reduced, and the service life of the whole hemodialysis machine can be prolonged.
Optionally, the preliminary determining whether the oscillogram includes an abnormal waveform segment based on image recognition includes the following steps:
constructing a basic recognition model based on a yolov3 model;
calling a plurality of historical abnormal waveform segments from a preset waveform database;
preprocessing the historical abnormal waveform section to obtain a preprocessed waveform section;
training the basic recognition model through the preprocessed waveform segment to obtain an abnormal recognition model;
and importing the oscillogram into the abnormal recognition model, and preliminarily judging whether the oscillogram contains an abnormal waveform section or not through the abnormal recognition model.
By adopting the technical scheme, after the basic waveform recognition model is built based on the yolov3 model, historical abnormal waveform segments can be called from the waveform database to train the basic recognition model so as to train the optimal abnormal recognition model, and then the abnormal recognition model is used for analyzing and judging the waveform diagram, so that whether the abnormal waveform segments containing suspected abnormal sounds exist in the waveform diagram can be preliminarily judged.
Optionally, the determining whether the hemodialysis pump is abnormal by combining the abnormal waveform segment and the abnormal frequency spectrogram includes the following steps:
judging whether the abnormal waveform section has a mutation point or not;
if the abrupt change point does not exist in the abnormal waveform segment, judging whether the blood turbine pump is abnormal or not according to the abnormal spectrogram;
if the abrupt change point exists in the abnormal waveform segment, calculating a kurtosis value of the abnormal waveform segment;
judging whether the kurtosis value is lower than a preset kurtosis threshold value or not;
if the kurtosis value is lower than the kurtosis threshold value, judging whether the blood turbine pump is abnormal or not according to the abnormal spectrogram;
if the kurtosis value is higher than the kurtosis threshold value, it is determined that the blood turbine pump is abnormal.
By adopting the technical scheme, whether instantaneous noise exists in the running process of the blood turbine pump is judged by judging whether a catastrophe point exists in the abnormal waveform section, if the catastrophe point exists, the blood turbine pump is possible to have abnormity such as part looseness, the kurtosis value of the abnormal waveform section needs to be further calculated, whether the blood turbine pump is abnormal is judged by comparing a preset kurtosis threshold value with the kurtosis value, and if the catastrophe point does not exist in the abnormal waveform section or the kurtosis value is lower than the kurtosis threshold value, the blood turbine pump needs to be further analyzed and judged according to the abnormal spectrogram.
Optionally, the formula for calculating the kurtosis value of the abnormal waveform segment is as follows:
in the formula, x i The signal value is the ith signal value in the abnormal waveform section, x is the signal mean value of the abnormal waveform section, and sigma is the standard deviation of all signals in the abnormal waveform section.
By adopting the technical scheme, if the blood turbine pump is in a normal operation state, the kurtosis value calculated by the formula is stabilized in a constant interval, and if the blood turbine pump has a pulse impact phenomenon caused by element breakage and aging during operation, the kurtosis value is increased.
Optionally, the determining whether the hemodialysis pump is abnormal according to the abnormal spectrogram includes the following steps:
calling a plurality of historical normal waveform segments from the waveform database;
adjusting the signal length of all the historical normal waveform segments by taking the abnormal waveform segment as a reference;
acquiring normal spectrograms of all the adjusted historical normal waveform segments, wherein the normal spectrograms and the abnormal spectrograms have the same spectrum resolution;
enveloping the abnormal spectrogram and all the normal spectrogram by using the amplitude as a characteristic value to obtain an abnormal enveloping chart and a plurality of normal enveloping charts;
dividing a reference frequency interval based on the normal enveloping graphs, and calculating the average interval area of the normal enveloping graphs in the reference frequency interval;
calculating the special interval area of the abnormal envelope map in the reference frequency interval;
judging whether the area of the special interval is smaller than the area of the average interval or not;
and if the area of the special interval is smaller than the area of the average interval, judging that the hemodialysis pump is abnormal.
By adopting the technical scheme, when the spectrogram is used for carrying out abnormity analysis, the normal spectrogram can be comprehensively analyzed, a reference frequency interval is defined, the average interval area between the envelope line and the horizontal axis is calculated, then the special interval area in the same interval in the abnormity envelope graph is calculated, and the special interval area and the average interval area are compared to judge whether the hemodialysis machine pump is abnormal or not.
Optionally, the method further includes the following steps after determining that the hemodialysis pump is abnormal:
marking the abnormal waveform segment with an abnormal label;
screening out the abnormal waveform section in the oscillogram and reserving a normal waveform in the oscillogram;
dividing the normal waveform into a plurality of normal waveform segments according to a preset standard signal length;
marking a plurality of normal waveform segments with normal labels respectively;
storing the abnormal waveform segment and all the normal waveform segments in the waveform database.
By adopting the technical scheme, if the blood turbine pump is judged to be abnormal, the acquired abnormal waveform section can be stored into the waveform database as an abnormal sample, so that the follow-up training of an abnormal recognition model is facilitated, the recognition accuracy of the abnormal recognition model is improved, and the normal waveform can also be stored into the waveform database as a normal sample, so that the recognition accuracy of follow-up spectrum abnormal analysis and recognition is facilitated.
Optionally, the acquiring an abnormal spectrogram of the abnormal waveform segment includes the following steps:
performing ADC processing on the abnormal waveform section to obtain an abnormal digital sample;
and carrying out fast Fourier transform on the abnormal digital sample based on a preset sampling frequency to obtain an abnormal frequency spectrogram.
By adopting the technical scheme, ADC processing can be carried out on the abnormal waveform section before fast Fourier transform is carried out, and the precision in fast Fourier transform is favorably improved.
In a second aspect, the present application further provides a blood turbine pump state warning system based on audio recognition, including a memory, a processor, and a program stored in the memory and executable on the processor, where the program is capable of being loaded and executed by the processor to implement the blood turbine pump state warning method based on audio recognition as described in the first aspect.
By adopting the technical scheme, through the calling of the program, the audio signal when the hemodialysis pump runs can be collected and the oscillogram can be drawn, whether the oscillogram contains an abnormal waveform section or not can be preliminarily judged according to the image recognition technology, if the oscillogram contains the abnormal waveform section, the abnormal waveform section and the abnormal frequency spectrogram can be combined for further analyzing and judging whether the hemodialysis pump is abnormal or not, and if the abnormal waveform section and the abnormal frequency spectrogram are detected, an abnormal alarm is sent out to remind a worker to overhaul the hemodialysis pump in time. Compared with manual regular maintenance in the related art, the operating state of the hemodialysis machine pump can be analyzed and judged only by the operating sound of the hemodialysis machine pump without frequently disassembling the hemodialysis machine, so that the faults of the hemodialysis machine pump can be reduced, and the service life of the whole hemodialysis machine can be prolonged.
To sum up, the application comprises the following beneficial technical effects:
through audio signal and the oscillogram of drawing when gathering the operation of hemodialysis machine pump, can be according to the preliminary judgement oscillogram of image identification technique whether contain unusual waveform section, if contain unusual waveform section, then can combine unusual waveform section and unusual spectrogram further analysis to judge the hemodialysis machine pump whether have unusually, if detect out unusually then send unusual warning to remind the staff in time to overhaul the hemodialysis machine pump. Compared with manual regular maintenance in the related art, the running state of the hemodialysis machine pump can be analyzed and judged only by the running sound of the hemodialysis machine pump, and the hemodialysis machine does not need to be frequently disassembled, so that the faults of the hemodialysis machine pump can be reduced, and the service life of the whole hemodialysis machine can be prolonged.
Drawings
Fig. 1 is a schematic flowchart of one implementation manner of a method for warning a state of a blood turbine pump based on audio recognition according to an embodiment of the present application.
Fig. 2 is a schematic flowchart of one implementation manner of a hemodialysis machine pump state warning method based on audio recognition according to an embodiment of the present application.
Fig. 3 is a schematic flowchart of one implementation manner of a hemodialysis machine pump state warning method based on audio recognition according to an embodiment of the present application.
Fig. 4 is a schematic flowchart of one implementation manner of a hemodialysis machine pump state warning method based on audio recognition according to an embodiment of the present application.
Fig. 5 is a flowchart illustrating an implementation manner of a hemodialysis machine pump state warning method based on audio recognition according to an embodiment of the present application.
Detailed Description
The present application is described in further detail below with reference to fig. 1 to 5.
The embodiment of the application discloses a hemodialysis machine pump state early warning method based on audio recognition.
Referring to fig. 1, the method for pre-warning the state of the hemodialysis machine pump based on audio recognition comprises the following steps:
s101, collecting audio signals when the hemodialysis machine pump runs.
The audio signal during the operation of the hemodialysis pump can be collected by a microphone arranged beside the hemodialysis pump in advance.
And S102, drawing a waveform diagram based on the audio signal.
The method comprises the steps of preprocessing an audio signal before drawing a oscillogram, wherein the preprocessing step comprises the steps of intercepting the audio signal based on preset duration, then denoising the intercepted audio signal, and drawing the oscillogram according to the denoised audio signal, wherein the oscillogram is a time sequence chart, and the horizontal axis is time and the vertical axis is amplitude.
S103, preliminarily judging whether the oscillogram contains an abnormal waveform section or not based on image identification, and if so, executing the step S104.
When the hemodialysis pump is in a normal operation state, the waveform of the waveform diagram is relatively stable, the amplitude is maintained below 0.6V, when the hemodialysis pump breaks down and other abnormal phenomena occur, the amplitude of the waveform in the waveform diagram has a sudden change phenomenon, and an abnormal waveform section with the overall average amplitude exceeding 0.6V can occur, and because the abnormal waveform section and the normal waveform section have relatively obvious difference on the graph, the abnormal waveform section in the waveform diagram can be preliminarily identified and judged in an image identification mode.
And S104, acquiring an abnormal frequency spectrogram of the abnormal waveform segment.
The method for acquiring the abnormal spectrogram of the abnormal waveform segment specifically comprises the following steps:
S104A, ADC processing is carried out on the abnormal waveform section to obtain an abnormal digital sample;
the ACD processing is filtering sampling processing.
S104B, performing fast Fourier transform on the abnormal digital sample based on a preset sampling frequency to obtain an abnormal frequency spectrogram.
The time domain can be converted into the frequency domain through fast Fourier transform, so that an abnormal spectrogram is obtained, wherein the horizontal axis in the spectrogram is frequency, the vertical axis in the spectrogram is frequency, and the amplitude in the spectrogram is amplitude.
And S105, judging whether the hemodialysis pump is abnormal or not by combining the abnormal waveform segment and the abnormal spectrogram, and executing the step S106 if the hemodialysis pump is abnormal.
And S106, sending an abnormal alarm.
The abnormity alarm comprises a buzzer arranged in the hemodialysis machine and is triggered to send out sound alarm, and alarm information is sent to a mobile terminal held by related staff.
The implementation principle of the embodiment is as follows:
the method comprises the steps of collecting audio signals when the blood turbine pump runs and drawing a oscillogram, preliminarily judging whether the oscillogram contains an abnormal waveform section or not according to an image recognition technology, if the oscillogram contains the abnormal waveform section, further analyzing and judging whether the blood turbine pump is abnormal or not by combining the abnormal waveform section and an abnormal frequency spectrogram, and if the abnormal waveform section is detected, giving an abnormal alarm to remind a worker to overhaul the blood turbine pump in time. Compared with manual regular maintenance in the related art, the running state of the hemodialysis machine pump can be analyzed and judged only by the running sound of the hemodialysis machine pump, and the hemodialysis machine does not need to be frequently disassembled, so that the faults of the hemodialysis machine pump can be reduced, and the service life of the whole hemodialysis machine can be prolonged.
In step S103 of the embodiment shown in fig. 1, a basic recognition model may be constructed based on yolov3, the model is trained to be an optimal model through the historical abnormal waveform segment, and the optimal model is used to perform preliminary recognition on the waveform diagram. This is explained in detail with reference to the embodiment shown in fig. 2.
Referring to fig. 2, the preliminary judgment of whether the oscillogram contains an abnormal waveform segment based on image recognition includes the following steps:
s201, constructing a basic recognition model based on the yolov3 model.
The yolov3 model is a target detection model, and can be used for constructing a basic identification model for identifying an abnormal waveform section by using the yolov3 model.
S202, calling a plurality of historical abnormal waveform segments from a preset waveform database.
The waveform database stores a large number of historical waveform sections in advance, the signal lengths of all the historical waveform sections are the same, the historical waveform sections comprise historical abnormal waveform sections and historical normal waveform sections, the historical abnormal waveform sections are waveforms of the hemodialysis pump in an abnormal working state obtained in a historical mode, and the historical normal waveform sections are waveforms of the hemodialysis pump in a normal working state obtained in the historical mode. The historical waveform section in the waveform database comprises a waveform processed by an audio signal collected through microphone history, and also comprises a waveform processed by an audio signal for searching the operation of the hemodialysis machine pumps of the same model through the Internet.
S203, preprocessing the historical abnormal waveform segment to obtain a preprocessed waveform segment.
The method mainly comprises the step of preprocessing a waveform image of a historical abnormal waveform segment, wherein the preprocessing step comprises denoising processing and image enhancement processing.
And S204, training a basic recognition model by preprocessing the waveform segment to obtain an abnormal recognition model.
The method comprises the steps of taking a large number of preprocessed waveform segments as training examples, importing the preprocessed waveform segments into a basic recognition model, training each parameter in the basic recognition model to an optimal parameter through model training, and obtaining the basic recognition model after training as an abnormal recognition model.
S205, the oscillogram is led into an abnormal recognition model, and whether the oscillogram contains an abnormal waveform section or not is preliminarily judged through the abnormal recognition model.
The method comprises the steps of identifying amplitude characteristics of a waveform diagram through an abnormal identification model, and circling a waveform section with an abrupt change of amplitude and a waveform section with an overall average amplitude exceeding a certain threshold value as abnormal waveform sections.
The implementation principle of the embodiment is as follows:
after a basic waveform recognition model is built based on the yolov3 model, historical abnormal waveform segments can be called from a waveform database to train the basic recognition model so as to train an optimal abnormal recognition model, and then the waveform diagram is analyzed and judged through the abnormal recognition model, so that whether the waveform diagram contains the abnormal waveform segments of suspected abnormal sounds or not can be preliminarily judged.
In step S105 of the embodiment shown in fig. 1, a determination is made according to the waveform characteristics and the kurtosis value of the abnormal waveform segment, and if it is not determined that the blood turbine pump is abnormal according to the abnormal waveform segment, the analysis determination is continued according to the abnormal spectrogram. This is explained in detail with reference to the embodiment shown in fig. 3.
Referring to fig. 3, the step of determining whether the hemodialysis pump is abnormal or not by combining the abnormal waveform segment and the abnormal spectrogram includes the following steps:
s301, judging whether a mutation point exists in the abnormal waveform section, and if the mutation point does not exist in the abnormal waveform section, executing the step S302; if there is a discontinuity in the abnormal waveform segment, step S303 is executed.
And judging through a preset mutation threshold, and if the amplitude value of any node in the abnormal waveform section exceeds the preset mutation threshold, indicating that a mutation point exists in the abnormal waveform section.
S302, judging whether the hemodialysis machine pump is abnormal or not according to the abnormal frequency spectrogram.
And S303, calculating the kurtosis value of the abnormal waveform segment.
The formula for calculating the kurtosis value of the abnormal waveform segment is as follows:
in the formula, x i Is the ith signal value in the abnormal waveform segment, x is the signal mean value of the abnormal waveform segment, and sigma is the standard deviation of all signals in the abnormal waveform segment. If the blood turbine pump is in a normal operation state, the kurtosis value calculated by the formula is stabilized in a constant interval, and if the blood turbine pump has a pulse impact phenomenon caused by element breakage and aging during operation, the kurtosis value is increased.
S304, judging whether the kurtosis value is lower than a preset kurtosis threshold, if so, executing the step S305; if the kurtosis value is higher than the kurtosis threshold, the step S306 is executed.
In the time domain, the kurtosis value is a common dimensionless characteristic parameter for abnormal signal analysis, is sensitive to impulse impact components, and can be used for judging the damage or aging of equipment components through analyzing the kurtosis value. The kurtosis value is usually maintained at about 3 in normal operation of the blood turbine pump, and when the blood turbine pump is in an abnormal operation state, the kurtosis value rises to 4 or more, so that the kurtosis threshold is usually set to 4. If the kurtosis value is equal to the kurtosis threshold, the same step S305 is performed.
S305, judging whether the blood vessel pump is abnormal or not according to the abnormal spectrogram.
S306, judging that the hemodialysis machine pump is abnormal.
The implementation principle of the embodiment is as follows:
whether instantaneous noise exists in the running process of the blood-turbine pump is judged by judging whether a catastrophe point exists in the abnormal waveform section, if the catastrophe point exists, the blood-turbine pump is possible to have abnormity such as part looseness, the kurtosis value of the abnormal waveform section needs to be further calculated, whether the blood-turbine pump is abnormal is judged by comparing a preset kurtosis threshold value with the kurtosis value, and if the catastrophe point does not exist in the abnormal waveform section or the kurtosis value is lower than the kurtosis threshold value, the blood-turbine pump needs to be further analyzed and judged according to the abnormal spectrogram.
In step S302 or step S305 of the embodiment shown in fig. 3, a normal envelope map and an abnormal envelope map are drawn according to the normal spectrogram and the abnormal spectrogram, a reference frequency interval is defined based on the normal spectrogram, an average interval area is calculated, a special interval area in the abnormal spectrogram is calculated, and whether the hemodialysis pump is abnormal or not can be determined by comparing the area sizes. This is explained in detail with reference to the embodiment shown in fig. 4.
Referring to fig. 4, the step of determining whether the hemodialysis pump is abnormal according to the abnormal spectrogram includes the following steps:
s401, a plurality of historical normal waveform segments are called from a waveform database.
The historical normal waveform segments in the waveform database are all marked with normal labels, so that a plurality of historical normal waveform segments can be retrieved from the waveform database based on the normal labels, and the plurality of historical normal waveform segments can be called out.
And S402, adjusting the signal lengths of all historical normal waveform segments by taking the abnormal waveform segment as a reference.
In order to facilitate subsequent frequency domain analysis, the lengths of the signals of the abnormal waveform segment and the historical normal waveform segment need to be unified.
And S403, acquiring the normal spectrogram of all the adjusted historical normal waveform segments.
The obtaining manner of the normal spectrogram can refer to steps S104A to S104B, and the spectrum resolutions of the normal spectrogram and the abnormal spectrogram are the same.
S404, enveloping the abnormal spectrogram and all the normal spectrograms by using the amplitude as a characteristic value to obtain an abnormal enveloping graph and a plurality of normal enveloping graphs.
The method comprises the steps of dividing N continuous points in an abnormal spectrogram into a group, dividing M groups in the whole abnormal spectrogram, taking the maximum value of the amplitude of the N points in each group as the characteristic value of the group, sequentially connecting the characteristic values of the M groups by using curves to obtain an abnormal envelope graph of the whole abnormal spectrogram, and obtaining a normal envelope graph by using the normal spectrogram in the same way.
And S405, defining a reference frequency interval based on the plurality of normal envelope maps, and calculating the average interval area of the plurality of normal envelope maps in the reference frequency interval.
By analyzing a plurality of normal envelope maps, a reference frequency interval can be defined, wherein in the frequency interval, the amplitude fluctuation in the normal envelope maps is large, and the amplitude fluctuation in the abnormal envelope maps is small. Calculating the average interval area of the plurality of normal envelope maps in the reference frequency interval is to calculate the base interval area of all the normal envelope maps, where the base interval area is the area of an area surrounded by the horizontal axis of the envelope maps, the envelope curve in the reference frequency interval, and two interval lines of the reference frequency interval, and assuming that the reference frequency interval is [1000hz,2000hz ], the two interval lines are x =1000 and x =2000. And after the basic interval area of each normal envelope image is calculated, calculating the average value of the basic interval areas of all the normal envelope images, namely the average interval area.
And S406, calculating the special interval area of the abnormal envelope graph in the reference frequency interval.
The method for calculating the area of the special section refers to the detailed description in step S405.
S407, judging whether the area of the special interval is smaller than the area of the average interval, if so, executing the step S408.
And if the area of the special interval is larger than or equal to the area of the average interval, judging that the hemodialysis machine pump is normal.
S408, judging that the hemodialysis machine pump is abnormal.
The implementation principle of the embodiment is as follows:
when the spectrogram is used for abnormal analysis, the normal spectrogram can be comprehensively analyzed, a reference frequency interval is defined, the average interval area between the envelope line and the horizontal axis is calculated, then the special interval area in the same interval in the abnormal envelope graph is calculated, and the size of the special interval area and the size of the average interval area are compared to judge whether the bleeding pump is abnormal or not.
After determining that there is an abnormality in the hemodialysis pump in step S306 of the embodiment shown in fig. 3 or step S408 of the embodiment shown in fig. 4, the acquired waveform map may be separated into an abnormal waveform segment and a normal waveform segment, and stored separately in the waveform database. This is illustrated in detail by the embodiment shown in fig. 5.
Referring to fig. 5, the method for determining the abnormality of the hemodialysis pump further includes the following steps:
s501, marking an abnormal label for the abnormal waveform segment.
S502, screening out abnormal waveform sections in the waveform diagram, and keeping normal waveforms in the waveform diagram.
Wherein, the abnormal waveform section in the oscillogram is cut off, and then the residual normal waveform is spliced and reserved.
And S503, dividing the normal waveform into a plurality of normal waveform segments according to the preset standard signal length.
And S504, marking normal labels for the plurality of normal waveform segments respectively.
And S505, storing the abnormal waveform segment and all the normal waveform segments in a waveform database.
The implementation principle of the embodiment is as follows:
if the blood turbine pump is judged to be abnormal, the acquired abnormal waveform section can be stored into the waveform database as an abnormal sample, subsequent training of the abnormal recognition model is facilitated, and therefore the recognition accuracy of the abnormal recognition model is improved.
The embodiment of the application also discloses a blood turbine pump state early warning system based on audio recognition, which comprises a memory, a processor and a program which is stored on the memory and can run on the processor, wherein the program can be loaded and executed by the processor to realize the blood turbine pump state early warning method based on audio recognition as shown in fig. 1 to 5.
The implementation principle of the embodiment is as follows:
through the calling of procedure, can gather the audio signal and draw the oscillogram when hemodialysis machine pump moves, can be according to image identification technique preliminary judgement oscillogram contains unusual waveform section, if contain unusual waveform section, then can combine unusual waveform section and unusual spectrogram further analysis to judge hemodialysis machine pump whether have unusually, if detect out unusually then send unusual warning to remind the staff in time to overhaul hemodialysis machine pump. Compared with manual regular maintenance in the related art, the operating state of the hemodialysis machine pump can be analyzed and judged only by the operating sound of the hemodialysis machine pump without frequently disassembling the hemodialysis machine, so that the faults of the hemodialysis machine pump can be reduced, and the service life of the whole hemodialysis machine can be prolonged.
The above embodiments are preferred embodiments of the present application, and the protection scope of the present application is not limited by the above embodiments, so: equivalent changes in structure, shape and principle of the present application shall be covered by the protection scope of the present application.
Claims (8)
1. A hemodialysis machine pump state early warning method based on audio recognition is characterized by comprising the following steps:
collecting audio signals when a hemodialysis machine pump runs;
drawing a waveform map based on the audio signal;
preliminarily judging whether the oscillogram contains an abnormal waveform section or not based on image identification;
if the oscillogram contains the abnormal waveform section, acquiring an abnormal frequency spectrogram of the abnormal waveform section;
judging whether the hemodialysis pump is abnormal or not by combining the abnormal waveform segment and the abnormal spectrogram;
and if the hemodialysis pump is abnormal, an abnormal alarm is given out.
2. The hemodialysis machine pump state early warning method based on audio recognition as claimed in claim 1, wherein the preliminary determination of whether the oscillogram contains abnormal waveform segments based on image recognition comprises the following steps:
constructing a basic recognition model based on the yolov3 model;
calling a plurality of historical abnormal waveform segments from a preset waveform database;
preprocessing the historical abnormal waveform section to obtain a preprocessed waveform section;
training the basic recognition model through the preprocessed waveform segment to obtain an abnormal recognition model;
and importing the oscillogram into the abnormal recognition model, and preliminarily judging whether the oscillogram contains an abnormal waveform section or not through the abnormal recognition model.
3. The method for warning the state of the blood turbine pump based on audio recognition as claimed in claim 2, wherein the step of determining whether the blood turbine pump is abnormal or not by combining the abnormal waveform segment and the abnormal spectrogram comprises the steps of:
judging whether a mutation point exists in the abnormal waveform section;
if the mutation points do not exist in the abnormal waveform segment, judging whether the blood turbine pump is abnormal or not according to the abnormal spectrogram;
if the abrupt change point exists in the abnormal waveform segment, calculating a kurtosis value of the abnormal waveform segment;
judging whether the kurtosis value is lower than a preset kurtosis threshold value or not;
if the kurtosis value is lower than the kurtosis threshold value, judging whether the blood turbine pump is abnormal or not according to the abnormal spectrogram;
if the kurtosis value is higher than the kurtosis threshold value, judging that the blood turbine pump is abnormal.
4. The hemodialysis machine pump state early warning method based on audio recognition as claimed in claim 3, wherein:
the formula for calculating the kurtosis value of the abnormal waveform segment is as follows:
in the formula, x i And the signal value is the ith signal value in the abnormal waveform section, x is the signal mean value of the abnormal waveform section, and sigma is the standard deviation of all signals in the abnormal waveform section.
5. The method for warning the state of the hemodialysis pump based on audio recognition as claimed in claim 3, wherein the step of determining whether the hemodialysis pump is abnormal according to the abnormal spectrogram comprises the steps of:
calling a plurality of historical normal waveform segments from the waveform database;
adjusting the signal length of all the historical normal waveform segments by taking the abnormal waveform segment as a reference;
acquiring normal spectrograms of all the adjusted historical normal waveform segments, wherein the normal spectrograms and the abnormal spectrograms have the same spectrum resolution;
enveloping the abnormal spectrogram and all the normal spectrogram by using the amplitude as a characteristic value to obtain an abnormal enveloping chart and a plurality of normal enveloping charts;
defining a reference frequency interval based on a plurality of normal envelope graphs, and calculating the average interval area of the plurality of normal envelope graphs in the reference frequency interval;
calculating the special interval area of the abnormal envelope map in the reference frequency interval;
judging whether the area of the special interval is smaller than the area of the average interval or not;
and if the area of the special interval is smaller than the area of the average interval, judging that the hemodialysis pump is abnormal.
6. The method for warning the state of the hemodialysis pump based on audio recognition of claim 5, wherein the step of determining the abnormality of the hemodialysis pump further comprises the following steps:
marking the abnormal waveform segment with an abnormal label;
screening out the abnormal waveform section in the waveform diagram, and reserving a normal waveform in the waveform diagram;
dividing the normal waveform into a plurality of normal waveform segments according to a preset standard signal length;
marking a plurality of normal waveform segments with normal labels respectively;
storing the abnormal waveform segment and all the normal waveform segments in the waveform database.
7. The method for warning the state of the hemodialysis machine pump based on audio recognition as claimed in claim 1, wherein the step of obtaining the abnormal spectrogram of the abnormal waveform segment comprises the steps of:
performing ADC processing on the abnormal waveform section to obtain an abnormal digital sample;
and carrying out fast Fourier transform on the abnormal digital sample based on a preset sampling frequency to obtain an abnormal spectrogram.
8. A blood turbine pump state warning system based on audio recognition, comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the program is capable of being loaded and executed by the processor to implement a blood turbine pump state warning method based on audio recognition according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211034978.0A CN115471827A (en) | 2022-08-26 | 2022-08-26 | Hemodialysis machine pump state early warning method and system based on audio recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211034978.0A CN115471827A (en) | 2022-08-26 | 2022-08-26 | Hemodialysis machine pump state early warning method and system based on audio recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115471827A true CN115471827A (en) | 2022-12-13 |
Family
ID=84368689
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211034978.0A Pending CN115471827A (en) | 2022-08-26 | 2022-08-26 | Hemodialysis machine pump state early warning method and system based on audio recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115471827A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117153365A (en) * | 2023-10-30 | 2023-12-01 | 中国人民解放军总医院第二医学中心 | Dialysis equipment running state early warning method and system based on audio identification |
-
2022
- 2022-08-26 CN CN202211034978.0A patent/CN115471827A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117153365A (en) * | 2023-10-30 | 2023-12-01 | 中国人民解放军总医院第二医学中心 | Dialysis equipment running state early warning method and system based on audio identification |
| CN117153365B (en) * | 2023-10-30 | 2024-03-12 | 中国人民解放军总医院第二医学中心 | Dialysis equipment running state early warning method and system based on audio identification |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110807024B (en) | Dynamic threshold anomaly detection method and system, storage medium and intelligent device | |
| WO2021196501A1 (en) | Big data-based fault monitoring method and system for industrial production line | |
| CN110160765B (en) | Impact characteristic identification method and system based on sound or vibration signal | |
| CN1836241A (en) | System and method for detecting signal artifacts | |
| CN109783903B (en) | Industrial water pipeline fault diagnosis method and system based on time sequence | |
| CN115471827A (en) | Hemodialysis machine pump state early warning method and system based on audio recognition | |
| CN109724802B (en) | Motor bearing weak fault diagnosis method based on spectrogram evaluation and optimization | |
| AU2018223214B2 (en) | Optical acoustic sensing system and method | |
| CN110995524B (en) | Flow data monitoring method and device, electronic equipment and computer readable medium | |
| CN111263284B (en) | Microphone fault detection method and device, electronic equipment and storage medium | |
| CN116292246B (en) | Fault monitoring method and system for vacuum pump | |
| CN112837799B (en) | Remote internet big data intelligent medical system based on block chain | |
| JP2023542371A (en) | Method and device for determining fetal cross section based on ultrasound video images | |
| CN117665935B (en) | Monitoring data processing method for broken rock mass supporting construction process | |
| EP3696543A3 (en) | Active oil debris monitor particle detection and monitoring system | |
| CN111831517A (en) | Data processing method and device, electronic equipment and storage medium | |
| US11560789B2 (en) | Method for pre-warning deformation of casing pipe according to change feature of b-value of hydraulic fracturing induced microseismicity | |
| CN109596354B (en) | Band-pass filtering method based on self-adaptive resonance frequency band identification | |
| KR101983603B1 (en) | Apparatus For Making A Predictive Diagnosis Of Nuclear Power Plant By Machine Learning And Augmented Reality | |
| CN111538755A (en) | Equipment operation state anomaly detection method based on normalized cross correlation and unit root detection | |
| CN115684833A (en) | Cable line degradation monitoring method and system based on harmonic diagnosis and storage medium | |
| CN111272368B (en) | Support and hanger degradation alarm method based on fundamental frequency data K-Means clustering | |
| CN117809696B (en) | Industrial equipment health assessment and fault prediction method and system based on acoustic analysis | |
| CN116863957B (en) | Method, device, equipment and storage medium for identifying operation state of industrial equipment | |
| CN116878728B (en) | Pressure sensor fault detection analysis processing system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |