CN113569490A - Label-free data enhancement method based on deep learning network - Google Patents
Label-free data enhancement method based on deep learning network Download PDFInfo
- Publication number
- CN113569490A CN113569490A CN202110901883.3A CN202110901883A CN113569490A CN 113569490 A CN113569490 A CN 113569490A CN 202110901883 A CN202110901883 A CN 202110901883A CN 113569490 A CN113569490 A CN 113569490A
- Authority
- CN
- China
- Prior art keywords
- charging
- battery
- medical
- data
- medical equipment
- 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
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000013135 deep learning Methods 0.000 title claims abstract description 14
- 238000007599 discharging Methods 0.000 claims abstract description 44
- 238000001514 detection method Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 13
- 230000032683 aging Effects 0.000 claims abstract description 10
- 238000012544 monitoring process Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims abstract description 7
- 238000007726 management method Methods 0.000 claims description 25
- 238000012360 testing method Methods 0.000 claims description 20
- 238000013500 data storage Methods 0.000 claims description 6
- 230000005059 dormancy Effects 0.000 claims description 5
- 230000002708 enhancing effect Effects 0.000 claims description 4
- 230000001502 supplementing effect Effects 0.000 claims description 4
- 238000009434 installation Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 3
- 230000005611 electricity Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K17/00—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations
- G06K17/0022—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations arrangements or provisious for transferring data to distant stations, e.g. from a sensing device
- G06K17/0029—Methods or arrangements for effecting co-operative working between equipments covered by two or more of main groups G06K1/00 - G06K15/00, e.g. automatic card files incorporating conveying and reading operations arrangements or provisious for transferring data to distant stations, e.g. from a sensing device the arrangement being specially adapted for wireless interrogation of grouped or bundled articles tagged with wireless record carriers
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Computer Networks & Wireless Communication (AREA)
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention discloses a label-free data enhancement method based on a deep learning network, which is applied to a medical equipment charging management system, and comprises the steps of obtaining a labeled sample set in the battery charging and discharging process, and training and generating a charging depth convolution judgment model and a discharging depth convolution judgment model according to the corresponding labeled sample set; and then acquiring a label-free sample set in a certain time period, wherein the label-free sample set comprises charging data of the medical charging cabinet and electric quantity use data of the medical equipment, and analyzing and processing the data in the label-free sample set according to a charging depth convolution and discharging depth convolution judgment model to generate a pre-classification group and a corresponding pre-judgment result. The invention provides a label-free data enhancement method applied to medical equipment charging management, which can realize relatively accurate preliminary prediction of medical equipment aging only by monitoring and analyzing charging data and discharging data without manually carrying out periodical disassembly detection on the medical equipment.
Description
Technical Field
The invention relates to the technical field of label-free data processing, in particular to a label-free data enhancement method based on a deep learning network.
Background
The self-training algorithm is to fit the pseudo-label predicted by one model to another previously learned model by training the model. For each machine learning item, data is the basis and is an integral part, and in supervised learning, data must be labeled in order to train the machine learning model. In recent years, the health assessment and prediction technology of mechanical equipment becomes a key technology of long-life and high-reliability mechanical equipment operation management, compared with the traditional equipment operation management technology, the health assessment and prediction technology has the advantages that early system performance decline can be found, the current and future operation health states of the equipment can be given, the regular maintenance of the existing equipment can be developed to the visual maintenance, the operation management cost can be greatly reduced, and the like. Few researchers pay attention to the aging detection problem of medical equipment, but actually, the existing medical management platform cannot completely realize automatic detection of data, the quantity of small and medium-sized medical equipment in a hospital is large, and if workers are required to be frequently assigned to carry out regular detection, the labor cost investment is high.
Disclosure of Invention
The invention aims to provide a label-free data enhancement method based on a deep learning network, which can realize relatively accurate preliminary estimation of product aging indexes only by monitoring and analyzing charging data and discharging data of medical equipment.
In order to achieve the purpose, the invention provides the following technical scheme: a label-free data enhancement method based on a deep learning network comprises the following steps:
step S1, respectively obtaining labeled sample sets of a battery charging process and a battery discharging process, and training and generating a charging depth convolution judgment model and a discharging depth convolution judgment model according to the corresponding labeled sample sets;
step S2, obtaining a label-free sample set in a certain time period, wherein the label-free sample set comprises charging data of a medical charging cabinet and electric quantity use data of medical equipment, analyzing and processing corresponding data in the label-free sample set according to the charging depth convolution judgment model and the discharging depth convolution judgment model generated in the step S1, and generating a pre-classification group and a corresponding pre-judgment result;
step S3, randomly extracting a plurality of batteries for each pre-classification group, and performing charge detection and discharge detection on each battery to generate a corresponding test result; comparing the test result with the pre-judgment result; if the difference between the test result and the predetermined result is within the first threshold, performing step S4, and if the difference between the test result and the predetermined result is greater than the first threshold, performing step S5;
s4, successfully verifying the charging depth convolution judgment model and the discharging depth convolution judgment model;
and step S5, supplementing the new test result data and the corresponding battery charging and discharging data into a labeled sample set, and retraining the charging depth convolution judgment model and the discharging depth convolution judgment model.
Preferably, the charging data of the medical charging cabinet comprises initial electric quantity information, and the medical charging cabinet is provided with a residual electric quantity detection unit for detecting residual electric quantity of the battery to be charged before charging operation and generating initial electric quantity information to be stored in a charging database of the medical charging cabinet; each battery is provided with a corresponding first RFID tag, the first RFID tag is stored with basic parameters of the battery, the basic parameters comprise battery capacity and battery model, each medical charging cabinet is provided with an RFID identifier, the RFID identifier is used for performing tag identification on the battery before the battery to be charged is charged, a data storage block of the battery is established in charging subdata according to the identification result, and charging data generated by the battery on the medical charging cabinet is stored in the data storage block.
Preferably, a remaining power prompting unit is arranged on the medical device and used for prompting the power when the remaining power is smaller than a second threshold, the medical device is configured with a power switching strategy, the power switching strategy comprises a power switching task which is started when power prompting information is received, after the power switching task is started, the medical device sends a rest command to the management platform, the medical device temporarily stops receiving a new execution task, and the execution of the power switching task is performed after the medical device completes a current work instruction.
Preferably, the charging data of the medical charging cabinet further includes a charging time and a charging current variation curve, the charging time and the charging current variation curve are both used for indicating performance parameters of charging equipment in the medical charging cabinet, the charging current variation curve is divided into at least three stages including a high-speed charging stage, a stable charging stage and a slow charging stage, and the charging data includes a time ratio of each stage and a start-end current value of each stage.
Preferably, the charging data of the medical charging cabinet further includes a total charging power supply amount, an electric quantity monitoring unit is arranged on the medical charging cabinet and used for monitoring electric quantity consumed in the charging process in real time, the electric quantity detection unit includes a detection resistor connected in series in a continuous power supply circuit and a voltage measuring meter connected in parallel with the detection resistor, a real-time current value is calculated according to a resistance value of the detection resistor and the real-time voltage value, and the real-time current value and time are integrated to generate the total charging power supply amount in the charging process; generating charging loss data by calculating the total charging power supply quantity, the initial electric quantity and the battery capacity; the charging loss data includes battery charging loss and charging equipment power supply loss.
Preferably, the management platform is configured with a data base, the medical equipment is provided with a second RFID tag, the battery replacement task comprises that the medical equipment automatically cruises and runs to a nearest medical charging cabinet, an RFID identifier on the medical charging cabinet carries out coding identification on the medical equipment, automatic battery replacement operation is carried out after the coding is verified to be error-free, a battery on the medical equipment is transferred to the medical charging cabinet for charging, a new fully charged battery is transferred to the medical equipment, the medical equipment sends a dormancy cancellation signal to the management platform after the fully charged battery is installed, and the task issued by the management platform is received again; after the battery replacement is finished, the medical charging cabinet transmits information that the medical equipment with the corresponding code is replaced by the battery with the corresponding code to the management platform, the management platform simultaneously receives a battery replacement finishing signal sent by the medical charging cabinet and a dormancy canceling signal sent by the medical equipment and issues an execution task to the corresponding medical equipment again, a plurality of data units matched with the medical equipment are stored in the data base, and historical battery replacement information of the medical equipment is stored in the data units; the power usage data of the medical device includes an effective usage duration of the new battery from installation on the medical device to a time period when the power prompt message is received.
Preferably, the labeled sample set in step S1 includes the charging data of the medical charging cabinet and the power usage data of the medical device, and the aging index, the battery charging efficiency index, the battery discharging efficiency index and the medical device power consumption index of the charging device in the corresponding medical charging cabinet.
Preferably, the charging depth convolution judgment model includes a medical charging cabinet charging sub-model and a battery charging sub-model, the step S1 is configured with a model generation strategy, the model generation strategy includes classifying the same type of data in the set of labeled samples, the classifying includes acquiring charging data of the same medical charging cabinet for charging different batteries when the medical charging cabinet charging sub-model is established, and acquiring charging data of the same battery for charging different medical charging cabinets when the battery charging sub-model is established.
Preferably, the model generation strategy comprises the step of allocating confidence degrees according to the charging data quantity of the medical charging cabinet with the label sample set and the charging quantity of the corresponding battery, wherein the confidence degree is higher when the charging data quantity is larger; and calculating according to the confidence coefficient and the charging loss generated in the charging process of the corresponding medical charging cabinet or the battery and other batteries or medical charging cabinets, distributing the ratio of the battery charging loss in the charging loss to the power supply loss of the charging equipment, and calculating the aging index and the battery charging efficiency index of the corresponding charging equipment.
Preferably, the medical equipment is provided with a discharge electric quantity detection unit, and the discharge electric quantity detection unit is connected in series with a power supply circuit of the battery and is used for detecting the total power supply quantity of the battery; generating discharge loss data by calculating a total power supply amount and an effective use time period of the medical equipment, wherein the power supply amounts consumed by different medical equipment per unit effective use time period are the same; the discharge loss data comprises battery discharge loss and medical equipment power consumption loss;
the classification processing comprises the steps of obtaining electric quantity use data of the same medical equipment which is charged by using different batteries and discharge data of the same battery which is discharged in different medical equipment;
the model generation strategy comprises the steps of distributing confidence degrees according to the quantity of electric quantity use data of the medical equipment with the label sample set, which are charged by using different batteries, and the quantity of discharge data of the corresponding batteries, which are discharged on different medical equipment, wherein the confidence degree is higher when the quantity of the data is more; and calculating according to the confidence coefficient and the discharge loss generated in the discharge process of the corresponding medical equipment or the battery and other batteries or medical equipment, distributing the proportion of the battery discharge loss and the power consumption loss of the medical equipment in the discharge loss, and calculating the corresponding battery discharge efficiency index and the corresponding medical equipment power consumption index.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a label-free data enhancement method based on a deep learning network, on one hand, the label-free data self-learning algorithm is introduced into the medical charging cabinet index judgment, the battery charging and discharging index judgment and the medical equipment power consumption loss index judgment of medical equipment, the medical charging cabinet, the battery and the medical equipment do not need to be manually and periodically detached and detected, and only the charging data and the discharging data need to be monitored and analyzed, so that the more accurate preliminary estimation of the product performance index can be realized. On the other hand, corresponding charge and discharge amount detection is set, a foundation is provided for data acquisition of charge data of the medical charging cabinet and electricity usage data of medical equipment, and a pre-judgment result is generated according to the charge depth convolution judgment model and the discharge depth convolution judgment model by using the new data indexes; and through spot check verification, the accuracy of the charging depth convolution judgment model and the accuracy of the discharging depth convolution judgment model are further continuously improved.
Drawings
Fig. 1 is a flow chart of a method for enhancing unlabeled data based on a deep learning network according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, a first embodiment provided by the present invention is a method for enhancing unlabeled data based on a deep learning network, including:
step S1, respectively obtaining labeled sample sets of a battery charging process and a battery discharging process, and training and generating a charging depth convolution judgment model and a discharging depth convolution judgment model according to the corresponding labeled sample sets;
step S2, obtaining a label-free sample set in a certain time period, wherein the label-free sample set comprises charging data of a medical charging cabinet and electric quantity use data of medical equipment, analyzing and processing corresponding data in the label-free sample set according to the charging depth convolution judgment model and the discharging depth convolution judgment model generated in the step S1, and generating a pre-classification group and a corresponding pre-judgment result;
step S3, randomly extracting a plurality of batteries for each pre-classification group, and performing charge detection and discharge detection on each battery to generate a corresponding test result; comparing the test result with the pre-judgment result;
step S31: judging whether the difference value between the test result and the pre-judgment result is within a first threshold value, if so, executing step S4, and if the difference value between the test result and the pre-judgment result is greater than the first threshold value, executing step S5;
s4, successfully verifying the charging depth convolution judgment model and the discharging depth convolution judgment model;
and step S5, supplementing the new test result data and the corresponding battery charging and discharging data into a labeled sample set, and retraining the charging depth convolution judgment model and the discharging depth convolution judgment model.
The algorithm of label-free data self-learning is introduced into medical charging cabinet index judgment, battery charging and discharging index judgment and medical equipment power consumption loss index judgment of the medical equipment, so that the medical charging cabinet, the battery and the medical equipment do not need to be manually and periodically detached for detection, and only charging data and discharging data need to be monitored and analyzed, and relatively accurate preliminary estimation of product performance indexes can be realized.
Preferably, the charging data of the medical charging cabinet comprises initial electric quantity information, and the medical charging cabinet is provided with a residual electric quantity detection unit for detecting residual electric quantity of the battery to be charged before charging operation and generating initial electric quantity information to be stored in a charging database of the medical charging cabinet; each battery is provided with a corresponding first RFID tag, the first RFID tag is stored with basic parameters of the battery, the basic parameters comprise battery capacity and battery model, each medical charging cabinet is provided with an RFID identifier, the RFID identifier is used for performing tag identification on the battery before the battery to be charged is charged, a data storage block of the battery is established in charging subdata according to the identification result, and charging data generated by the battery on the medical charging cabinet is stored in the data storage block.
Preferably, the medical device is provided with a remaining power prompting unit for prompting the power when the remaining power is smaller than a second threshold, the medical device is configured with a power swapping strategy, the power swapping strategy includes starting a power swapping task prompt when power prompting information is received, and execution of the power swapping task is performed after the medical device completes a current work instruction (for example, after data after collection and calculation processing is transmitted to a management platform).
The medical equipment can be small and medium-sized medical equipment or instruments such as a monitor, an electrocardiograph, a B-ultrasonic instrument, a breathing machine and the like which are electrically connected with the management platform and send processing data to the management platform.
Preferably, the charging data of the medical charging cabinet further includes a charging time and a charging current variation curve, the charging time and the charging current variation curve are both used for indicating performance parameters of charging equipment in the medical charging cabinet, the charging current variation curve is divided into at least three stages including a high-speed charging stage, a stable charging stage and a slow charging stage, and the charging data includes a time ratio of each stage and a start-end current value of each stage.
Preferably, the charging data of the medical charging cabinet further includes a total charging power supply amount, an electric quantity monitoring unit is arranged on the medical charging cabinet and used for monitoring electric quantity consumed in the charging process in real time, the electric quantity detection unit includes a detection resistor connected in series in a continuous power supply circuit and a voltage measuring meter connected in parallel with the detection resistor, a real-time current value is calculated according to a resistance value of the detection resistor and the real-time voltage value, and the real-time current value and time are integrated to generate the total charging power supply amount in the charging process; generating charging loss data by calculating the total charging power supply quantity, the initial electric quantity and the battery capacity; the charging loss data includes battery charging loss and charging equipment power supply loss.
Preferably, the management platform is configured with a data bank, the medical device is provided with a second RFID tag, the power exchange task includes that the medical device is moved to a corresponding medical charging cabinet, an RFID identifier on the medical charging cabinet performs encoding identification on the medical device, and after the encoding is verified to be correct, carrying out automatic battery replacement operation, transferring the battery on the medical equipment to a medical charging cabinet for charging, transferring a new fully charged battery to the medical equipment, after the battery replacement is finished, the medical charging cabinet transmits information that the medical equipment with the corresponding code is replaced by the battery with the corresponding code to the management platform, the management platform simultaneously receives a power exchange completion signal sent by the medical charging cabinet and a dormancy cancellation signal sent by the medical equipment and issues an execution task to the corresponding medical equipment again, a plurality of data units matched with the medical equipment are stored in the data total bank, and historical power exchange information of the medical equipment is stored in the data units; the power usage data of the medical device includes an effective usage duration of the new battery from installation on the medical device to a time period when the power prompt message is received. The effective use duration can be provided through the management platform.
Preferably, the labeled sample set in step S1 includes the charging data of the medical charging cabinet and the power usage data of the medical device, and the aging index, the battery charging efficiency index, the battery discharging efficiency index and the medical device power consumption index of the charging device in the corresponding medical charging cabinet. The aging index of the charging device is used to indicate the loss of electric energy generated by the charging device during charging, and if the loss is too large, the charging device needs to be replaced or maintained. The battery charging efficiency index and the battery discharging efficiency index respectively represent the increase of the battery along with the service time, the self-consumed electric quantity in the charging process and the discharging process or whether the self-discharging phenomenon exists in the discharging process, and if the self-consumed electric quantity is too much, the battery needs to be replaced. The power consumption index of the medical equipment indicates the change of the power consumption performance of the medical equipment along with the increase of the service time of the medical equipment, and if a certain threshold value is exceeded, the maintenance is needed or the service time of the corresponding medical equipment is reduced.
Preferably, the charging depth convolution judgment model includes a medical charging cabinet charging sub-model and a battery charging sub-model, the step S1 is configured with a model generation strategy, the model generation strategy includes classifying the same type of data in the set of labeled samples, the classifying includes acquiring charging data of the same medical charging cabinet for charging different batteries when the medical charging cabinet charging sub-model is established, and acquiring charging data of the same battery for charging different medical charging cabinets when the battery charging sub-model is established.
Preferably, the model generation strategy comprises the step of allocating confidence degrees according to the charging data quantity of the medical charging cabinet with the label sample set and the charging quantity of the corresponding battery, wherein the confidence degree is higher when the charging data quantity is larger; and calculating according to the confidence coefficient and the charging loss generated in the charging process of the corresponding medical charging cabinet or the battery and other batteries or medical charging cabinets, distributing the ratio of the battery charging loss in the charging loss to the power supply loss of the charging equipment, and calculating the aging index and the battery charging efficiency index of the corresponding charging equipment.
Preferably, the medical equipment is provided with a discharge electric quantity detection unit, and the discharge electric quantity detection unit is connected in series with a power supply circuit of the battery and is used for detecting the total power supply quantity of the battery; generating discharge loss data by calculating a total power supply amount and an effective use time period of the medical equipment, wherein the power supply amounts consumed by different medical equipment per unit effective use time period are the same; the discharge loss data comprises battery discharge loss and medical equipment power consumption loss;
the classification processing comprises the steps of obtaining electric quantity use data of the same medical equipment which is charged by using different batteries and discharge data of the same battery which is discharged in different medical equipment;
the model generation strategy comprises the steps of distributing confidence degrees according to the quantity of electric quantity use data of the medical equipment with the label sample set, which are charged by using different batteries, and the quantity of discharge data of the corresponding batteries, which are discharged on different medical equipment, wherein the confidence degree is higher when the quantity of the data is more; and calculating according to the confidence coefficient and the discharge loss generated in the discharge process of the corresponding medical equipment or the battery and other batteries or medical equipment, distributing the proportion of the battery discharge loss and the power consumption loss of the medical equipment in the discharge loss, and calculating the corresponding battery discharge efficiency index and the corresponding medical equipment power consumption index.
The working principle is as follows: firstly, respectively acquiring labeled sample sets in a battery charging process and a battery discharging process, and training and generating a charging depth convolution judgment model and a discharging depth convolution judgment model according to the corresponding labeled sample sets; then acquiring a label-free sample set in a certain fixed time period, wherein the judgment accuracy is influenced by too long time intervals (the service life of a new medical equipment battery is at least more than 10 years), the label-free sample set comprises charging data of a medical charging cabinet and electric quantity use data of medical equipment, analyzing and processing corresponding data in the label-free sample set according to a charging depth convolution judgment model and a discharging depth convolution judgment model, and generating a pre-classification group and a corresponding pre-judgment result; randomly extracting a plurality of batteries for each pre-classification group, and performing charge detection and discharge detection on each battery to generate a corresponding test result; comparing the test result with the pre-judgment result; and if the difference value between the test result and the pre-judgment result is within a first threshold value, successfully verifying the charging depth convolution judgment model and the discharging depth convolution judgment model, and if the difference value between the test result and the pre-judgment result is greater than the first threshold value, supplementing new test result data and corresponding battery charging and discharging data into a labeled sample set and retraining the charging depth convolution judgment model and the discharging depth convolution judgment model.
According to the method for enhancing the unlabelled data based on the deep learning network, the unlabelled data self-learning algorithm is introduced into the medical charging cabinet index judgment, the battery charging and discharging index judgment and the medical equipment power consumption loss index judgment of the medical equipment, the medical charging cabinet, the battery and the medical equipment do not need to be manually and periodically detached and detected, and only the charging data and the discharging data need to be monitored and analyzed, so that the relatively accurate preliminary estimation of the product performance index can be realized. On the other hand, corresponding charge and discharge amount detection is set, a foundation is provided for data acquisition of charge data of the medical charging cabinet and electricity usage data of medical equipment, and a pre-judgment result is generated according to the charge depth convolution judgment model and the discharge depth convolution judgment model by using the new data indexes; and through spot check verification, the accuracy of the charging depth convolution judgment model and the accuracy of the discharging depth convolution judgment model are further continuously improved.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
Claims (10)
1. A label-free data enhancement method based on a deep learning network is applied to a medical equipment charging management system, the medical equipment charging management system comprises a plurality of medical equipment, a plurality of medical charging cabinets and a management platform, and a battery for supplying power to the medical equipment is installed in the medical equipment, and the method is characterized by comprising the following steps:
step S1, respectively obtaining labeled sample sets of a battery charging process and a battery discharging process, and training and generating a charging depth convolution judgment model and a discharging depth convolution judgment model according to the corresponding labeled sample sets;
step S2, obtaining a label-free sample set in a certain time period, wherein the label-free sample set comprises charging data of a medical charging cabinet and electric quantity use data of medical equipment, analyzing and processing corresponding data in the label-free sample set according to the charging depth convolution judgment model and the discharging depth convolution judgment model generated in the step S1, and generating a pre-classification group and a corresponding pre-judgment result;
step S3, randomly extracting a plurality of batteries for each pre-classification group, and performing charge detection and discharge detection on each battery to generate a corresponding test result; comparing the test result with the pre-judgment result; if the difference between the test result and the predetermined result is within the first threshold, performing step S4, and if the difference between the test result and the predetermined result is greater than the first threshold, performing step S5;
s4, successfully verifying the charging depth convolution judgment model and the discharging depth convolution judgment model;
and step S5, supplementing the new test result data and the corresponding battery charging and discharging data into a labeled sample set, and retraining the charging depth convolution judgment model and the discharging depth convolution judgment model.
2. The method for the unlabeled data enhancement based on the deep learning network of claim 1, wherein: the charging data of the medical charging cabinet comprises initial electric quantity information, and a residual electric quantity detection unit is arranged on the medical charging cabinet and is used for detecting the residual electric quantity of the battery to be charged before charging operation is carried out on the battery to be charged and generating initial electric quantity information to be stored in a charging database of the medical charging cabinet; each battery is provided with a corresponding first RFID tag, the first RFID tag is stored with basic parameters of the battery, the basic parameters comprise battery capacity and battery model, each medical charging cabinet is provided with an RFID identifier, the RFID identifier is used for performing tag identification on the battery before the battery to be charged is charged, a data storage block of the battery is established in charging subdata according to the identification result, and charging data generated by the battery on the medical charging cabinet is stored in the data storage block.
3. The method of claim 2, wherein the method comprises the following steps: the medical equipment is provided with a residual electric quantity prompting unit used for prompting the electric quantity when the residual electric quantity is smaller than a second threshold value, the medical equipment is configured with a battery replacement strategy, the battery replacement strategy comprises a battery replacement task prompting started when electric quantity prompting information is received, and the execution of the battery replacement task is carried out after the medical equipment completes the current working instruction.
4. The method according to claim 3, wherein the method comprises the following steps: the charging data of the medical charging cabinet further comprises a charging time and a charging current change curve, wherein the charging time and the charging current change curve are used for indicating performance parameters of charging equipment in the medical charging cabinet, the charging current change curve is divided into at least three stages including a high-speed charging stage, a stable charging stage and a slow charging stage, and the charging data comprises a time ratio of each stage and a starting current value and a finishing current value of each stage.
5. The method of claim 4, wherein the method comprises: the charging data of the medical charging cabinet also comprises total charging power supply quantity, an electric quantity monitoring unit is arranged on the medical charging cabinet and used for monitoring the electric quantity consumed in the charging process in real time, the electric quantity detecting unit comprises a detecting resistor connected in series in a power supply circuit and a voltage measuring meter connected with the detecting resistor in parallel, a real-time current value is calculated according to the resistance value and the real-time voltage value of the detecting resistor, and the real-time current value and time are integrated to generate the total power supply quantity in the charging process; generating charging loss data by calculating the total charging power supply quantity, the initial electric quantity and the battery capacity; the charging loss data includes battery charging loss and charging equipment power supply loss.
6. The method of claim 4, wherein the method comprises: the management platform is configured with a data base, a second RFID label is arranged on the medical equipment, the battery replacement task comprises that the medical equipment automatically cruises and runs to a nearest medical charging cabinet, an RFID identifier on the medical charging cabinet carries out code identification on the medical equipment, automatic battery replacement operation is carried out after the code is verified to be error-free, a battery on the medical equipment is transferred to the medical charging cabinet for charging, a new fully charged battery is transferred to the medical equipment, the medical equipment sends a signal for canceling the dormancy to the management platform after the fully charged battery is installed, and the task issued by the management platform is received again; after the battery replacement is finished, the medical charging cabinet transmits information that the medical equipment with the corresponding code is replaced by the battery with the corresponding code to the management platform, the management platform simultaneously receives a battery replacement finishing signal sent by the medical charging cabinet and a dormancy canceling signal sent by the medical equipment and issues an execution task to the corresponding medical equipment again, a plurality of data units matched with the medical equipment are stored in the data base, and historical battery replacement information of the medical equipment is stored in the data units; the power usage data of the medical device includes an effective usage duration of the new battery from installation on the medical device to a time period when the power prompt message is received.
7. The method for label-free data enhancement based on the deep learning network as claimed in any one of claims 1-6, wherein: the labeled sample set in step S1 includes the charging data of the medical charging cabinet and the power usage data of the medical device, and the aging index, the battery charging efficiency index, the battery discharging efficiency index and the medical device power consumption index of the charging device in the corresponding medical charging cabinet.
8. The method according to claim 7, wherein the method comprises: the charging depth convolution judgment model comprises a medical charging cabinet charging sub-model and a battery charging sub-model, the step S1 is configured with a model generation strategy, the model generation strategy comprises classification processing of the same type of data in a label sample set, the classification processing comprises acquiring charging data of the same medical charging cabinet for charging different batteries when the medical charging cabinet charging sub-model is constructed, and acquiring charging data of the same battery for charging different medical charging cabinets when the battery charging sub-model is constructed.
9. The method according to claim 8, wherein the method comprises the following steps: the model generation strategy comprises the steps of distributing confidence degrees according to the quantity of charging data of the medical charging cabinet with the label sample set and the charging quantity of the corresponding battery, wherein the confidence degree is higher when the quantity of the charging data is larger; and calculating according to the confidence coefficient and the charging loss generated in the charging process of the corresponding medical charging cabinet or the battery and other batteries or medical charging cabinets, distributing the ratio of the battery charging loss in the charging loss to the power supply loss of the charging equipment, and calculating the aging index and the battery charging efficiency index of the corresponding charging equipment.
10. The method according to claim 9, wherein the method for enhancing unlabeled data based on deep learning network is characterized in that: the medical equipment is provided with a discharge electric quantity detection unit which is connected in series with a power supply circuit of the battery and is used for detecting the total power supply quantity of the battery; generating discharge loss data by calculating a total power supply amount and an effective use time period of the medical equipment, wherein the power supply amounts consumed by different medical equipment per unit effective use time period are the same; the discharge loss data comprises battery discharge loss and medical equipment power consumption loss;
the classification processing comprises the steps of obtaining electric quantity use data of the same medical equipment which is charged by using different batteries and discharge data of the same battery which is discharged in different medical equipment;
the model generation strategy comprises the steps of distributing confidence degrees according to the quantity of electric quantity use data of the medical equipment with the label sample set, which are charged by using different batteries, and the quantity of discharge data of the corresponding batteries, which are discharged on different medical equipment, wherein the confidence degree is higher when the quantity of the data is more; and calculating according to the confidence coefficient and the discharge loss generated in the discharge process of the corresponding medical equipment or the battery and other batteries or medical equipment, distributing the proportion of the battery discharge loss and the power consumption loss of the medical equipment in the discharge loss, and calculating the corresponding battery discharge efficiency index and the corresponding medical equipment power consumption index.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110901883.3A CN113569490A (en) | 2021-08-06 | 2021-08-06 | Label-free data enhancement method based on deep learning network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110901883.3A CN113569490A (en) | 2021-08-06 | 2021-08-06 | Label-free data enhancement method based on deep learning network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113569490A true CN113569490A (en) | 2021-10-29 |
Family
ID=78170712
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110901883.3A Pending CN113569490A (en) | 2021-08-06 | 2021-08-06 | Label-free data enhancement method based on deep learning network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113569490A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080312852A1 (en) * | 2004-11-08 | 2008-12-18 | Koninklijke Philips Electronics N.V. | Wireless Battery Status Management for Medical Devices |
| CN103227350A (en) * | 2013-04-17 | 2013-07-31 | 深圳市科曼医疗设备有限公司 | Intelligent battery management system and method for medical equipment |
| CN107221986A (en) * | 2017-07-10 | 2017-09-29 | 方奕海 | A kind of Medical Devices charging and management control system |
| CN111414942A (en) * | 2020-03-06 | 2020-07-14 | 重庆邮电大学 | Remote sensing image classification method based on active learning and convolutional neural network |
| CN112085071A (en) * | 2020-08-21 | 2020-12-15 | 广东电网有限责任公司广州供电局 | Power distribution room equipment fault analysis and pre-judgment method and device based on edge calculation |
| CN112467892A (en) * | 2020-10-22 | 2021-03-09 | 宁波大学 | Multi-antenna-array-based charging system for implantable medical electronic device |
| CN112986830A (en) * | 2021-04-22 | 2021-06-18 | 湖北工业大学 | Lithium battery capacity estimation method based on convolution time memory neural network |
-
2021
- 2021-08-06 CN CN202110901883.3A patent/CN113569490A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080312852A1 (en) * | 2004-11-08 | 2008-12-18 | Koninklijke Philips Electronics N.V. | Wireless Battery Status Management for Medical Devices |
| CN103227350A (en) * | 2013-04-17 | 2013-07-31 | 深圳市科曼医疗设备有限公司 | Intelligent battery management system and method for medical equipment |
| CN107221986A (en) * | 2017-07-10 | 2017-09-29 | 方奕海 | A kind of Medical Devices charging and management control system |
| CN111414942A (en) * | 2020-03-06 | 2020-07-14 | 重庆邮电大学 | Remote sensing image classification method based on active learning and convolutional neural network |
| CN112085071A (en) * | 2020-08-21 | 2020-12-15 | 广东电网有限责任公司广州供电局 | Power distribution room equipment fault analysis and pre-judgment method and device based on edge calculation |
| CN112467892A (en) * | 2020-10-22 | 2021-03-09 | 宁波大学 | Multi-antenna-array-based charging system for implantable medical electronic device |
| CN112986830A (en) * | 2021-04-22 | 2021-06-18 | 湖北工业大学 | Lithium battery capacity estimation method based on convolution time memory neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110503338B (en) | Ubiquitous power Internet of things monitoring method | |
| CN105938183B (en) | A kind of intelligent monitor system in prediction UPS battery service life | |
| CN108776276A (en) | Multiplexing electric abnormality detection method and system | |
| CN115330275B (en) | Echelon utilization method and device for retired battery | |
| CN116646965B (en) | Photovoltaic energy storage charging and discharging integrated management system | |
| CN109409444A (en) | A kind of method of discrimination of the polynary electric network fault type based on prior probability | |
| CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
| CN114865668A (en) | Energy storage scheduling support evaluation method | |
| CN117691645A (en) | Energy storage system for intelligent micro-grid | |
| CN115514068A (en) | Cell pressure difference optimization method for gradient utilization of lithium battery | |
| CN115456034A (en) | Automatic identification and monitoring method and system for electric bicycle charging | |
| CN117347869B (en) | Data analysis method and device for energy storage battery management system, electronic equipment and medium | |
| CN116885763B (en) | Energy management device and method suitable for distributed energy storage system | |
| CN113541306A (en) | Green energy low-carbon efficiency monitoring system and method | |
| CN206542219U (en) | A kind of MW class energy-accumulating power station battery management system | |
| CN113569490A (en) | Label-free data enhancement method based on deep learning network | |
| CN117371788A (en) | Electric power marketing risk early warning method and system | |
| CN115742781A (en) | Battery charging method and device, electronic equipment and storage medium thereof | |
| CN113064075B (en) | Motor service life prediction method based on edge calculation and deep learning | |
| CN114019393A (en) | Data-based battery capacity prediction method and system | |
| Jincheng et al. | Fault prediction of intelligent electricity meter based on multi-classification machine learning model | |
| CN114487848A (en) | Method and device for calculating state of storage battery | |
| CN110266003A (en) | House intelligent distribution system | |
| CN110912270A (en) | Distribution automation intelligence feeder terminal | |
| CN113363870B (en) | Intelligent operation and maintenance method for insulator based on autonomous controllable chip |
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 |