Intelligent sterilization cabinet fault alarm method and device based on sensor monitoring

Abstract

The application discloses a sensor monitoring-based intelligent disinfection cabinet fault alarm method and device, and relates to the technical field of fault early warning.The method comprises the following steps: acquiring a single monitoring data sequence of a disinfection cabinet through sensor monitoring according to a single monitoring index; acquiring a first water pump fault probability by predicting the single monitoring data sequence by using a fault early warning submodel; if the first water pump fault probability is greater than a first fault probability threshold, acquiring a multi-dimensional monitoring data sequence according to multi-dimensional monitoring indexes; acquiring a second water pump fault probability by predicting the multi-dimensional monitoring data sequence by using a fault early warning parent model; and if the second water pump fault probability is greater than a second fault probability threshold, performing water pump fault early warning of the disinfection cabinet.The technical problem of insufficient accuracy of disinfection cabinet fault early warning in the prior art is solved, and the technical effect of improving the accuracy of disinfection cabinet fault early warning is achieved by adopting a double early warning mechanism combining single monitoring index preliminary screening and multi-dimensional monitoring index review.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of fault early warning, in particular to an intelligent sterilization cabinet fault alarm method and device based on sensor monitoring. BACKGROUND

[0002] As a common health protection device in daily life and commercial environment, a sterilization cabinet is widely used in families, catering, medical treatment and the like. In order to guarantee the sterilization effect and the safety of the use of the device, the sterilization cabinet is usually integrated with multiple sterilization functions such as ultraviolet rays, high temperature and ozone, and the operation process thereof depends on the cooperative work of various internal motors, heating units and water pumps and the like. As an important component, the water pump is responsible for the water circulation and the steam generation process, and once a fault occurs, it may lead to incomplete sterilization, damage of the device and even safety hazards. However, the existing fault early warning mechanism is mostly based on static threshold setting and depends on a single sensor parameter for judgment, and it is difficult to effectively adapt to the complex and changeable operation environment, and problems such as false alarm, missed alarm or delayed response are prone to occur, and it is difficult to meet the intelligent and high-reliability device operation and maintenance requirements. SUMMARY

[0003] The application provides an intelligent sterilization cabinet fault alarm method and device based on sensor monitoring, which solves the technical problem of insufficient accuracy of sterilization cabinet fault early warning in the prior art.

[0004] In a first aspect, the application provides an intelligent sterilization cabinet fault alarm method based on sensor monitoring, which comprises the following steps:

[0005] According to a single monitoring index, a single monitoring data sequence of the sterilization cabinet in a preset time zone is obtained through sensor monitoring; a first water pump fault probability in a future time window is predicted and obtained according to the single monitoring data sequence by using a fault early warning sub-model; if the first water pump fault probability is greater than a first fault probability threshold, a multi-dimensional monitoring data sequence of the sterilization cabinet in a preset time period is obtained through sensor monitoring according to a multi-dimensional monitoring index; a second water pump fault probability in a future time window is predicted and obtained according to the multi-dimensional monitoring data sequence by using a fault early warning parent model; if the first water pump fault probability is greater than the first fault probability threshold, a first fault probability overflow value is calculated; a branch selection number P is obtained by analyzing the first fault probability overflow value; P fault early warning branches are randomly selected from K fault early warning branches of the fault early warning parent model, P fault probabilities are predicted according to the multi-dimensional monitoring data sequence, and a second water pump fault probability is obtained after mean value calculation; if the second water pump fault probability is greater than a second fault probability threshold, a water pump fault early warning of the sterilization cabinet is performed, wherein the second fault probability threshold is greater than the first fault probability threshold.

[0006] The second aspect of the present application provides a smart sterilization cabinet fault alarm device based on sensor monitoring, which comprises:

[0007] The monitoring module: according to a single monitoring index, a single monitoring data sequence of the sterilization cabinet in a preset time zone is obtained through sensor monitoring; the first fault prediction module: a first water pump fault probability in a future time window is predicted and obtained according to the single monitoring data sequence by using a fault early warning sub-model; the judgment module: if the first water pump fault probability is greater than a first fault probability threshold, a multi-dimensional monitoring data sequence of the sterilization cabinet in a preset time period is obtained through sensor monitoring according to a multi-dimensional monitoring index; the second fault prediction module: a second water pump fault probability in a future time window is predicted and obtained according to the multi-dimensional monitoring data sequence by using a fault early warning parent model; the early warning module: if the second water pump fault probability is greater than a second fault probability threshold, a water pump fault early warning of the sterilization cabinet is performed, wherein the second fault probability threshold is greater than the first fault probability threshold.

[0008] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] Firstly, a single monitoring data sequence of the sterilization cabinet in a preset time zone is obtained through sensor monitoring according to a single monitoring index. Then, a first water pump fault probability in a future time window is predicted and obtained according to the single monitoring data sequence by using a fault early warning sub-model. If the first water pump fault probability is greater than a first fault probability threshold, a multi-dimensional monitoring data sequence of the sterilization cabinet in a preset time period is obtained through sensor monitoring according to a multi-dimensional monitoring index. Then, a second water pump fault probability in a future time window is predicted and obtained according to the multi-dimensional monitoring data sequence by using a fault early warning parent model; if the first water pump fault probability is greater than the first fault probability threshold, a first fault probability overflow value is calculated; a branch selection number P is analyzed according to the first fault probability overflow value; P fault early warning branches are randomly selected in K fault early warning branches of the fault early warning parent model, P fault probabilities are predicted according to the multi-dimensional monitoring data sequence, and a second water pump fault probability is obtained after mean calculation. If the second water pump fault probability is greater than a second fault probability threshold, a water pump fault early warning of the sterilization cabinet is performed, wherein the second fault probability threshold is greater than the first fault probability threshold. The technical problem of insufficient accuracy of sterilization cabinet fault early warning in the prior art is solved, a double early warning mechanism combining single monitoring index preliminary screening and multi-dimensional monitoring index review is adopted, and the technical effect of improving the accuracy of sterilization cabinet fault early warning is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0011] Figure 1 A flowchart of a sensor monitoring-based intelligent disinfection cabinet fault alarm method provided by the embodiments of the present application is shown.

[0012] Figure 2 A structural diagram of a sensor monitoring-based intelligent disinfection cabinet fault alarm device provided by the embodiments of the present application is shown.

[0013] Legend: monitoring module 11, first fault prediction module 12, judgment module 13, second fault prediction module 14, and early warning module 15. DETAILED DESCRIPTION

[0014] The present application provides a sensor monitoring-based intelligent disinfection cabinet fault alarm method and device, which solves the technical problem of insufficient fault early warning accuracy of the prior art.

[0015] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without any creative effort fall within the scope of the present application.

[0016] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units need not be limited to only those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to the process, method, product or device.

[0017] Embodiment one, as shown in the present application provides a sensor monitoring-based intelligent disinfection cabinet fault alarm method, wherein the method comprises: Figure 1

[0018] According to a single monitoring index, a single monitoring data sequence of the disinfection cabinet in a preset time zone is obtained through sensor monitoring.

[0019] According to a preset water pump fault monitoring target, one monitoring index with the highest correlation degree with the water pump fault in historical data is selected from a plurality of monitoring parameters that can represent the running state of the water pump as the single monitoring index. ​

[0020] After the single monitoring index is determined, the system periodically monitors the single index according to a preset sampling period and monitoring time zone. Specifically, if the fault correlation degree of the current parameter is the highest, the current is selected as the single monitoring index; in each monitoring period (e.g., once per second), real-time current data is obtained through the current sensor deployed inside the disinfection cabinet or on the water pump pipeline, and is recorded as the monitoring value at the current time. The entire monitoring process lasts for a preset time zone, e.g., 5 minutes, and the system collects current values at several time points and forms a time sequence, i.e., a single monitoring data sequence, in chronological order.

[0021] Further, configuring the single monitoring index includes:

[0022] obtaining multi-dimensional monitoring indexes for water pump fault analysis, wherein the multi-dimensional monitoring indexes at least include water pressure parameters, water level parameters, and current parameters; based on historical water pump fault monitoring records, performing fault correlation analysis on the water pressure parameters, water level parameters, and current parameters respectively, and setting a single monitoring index according to the analysis results.

[0023] The system obtains multi-dimensional monitoring indexes for water pump fault analysis, which at least include water pressure parameters, water level parameters, and current parameters. The system calls stored historical water pump operation data and fault records to construct a water pump fault monitoring history database, which includes multi-dimensional monitoring parameter data (e.g., time sequences of water pressure, water level, and current) in multiple time periods and label information (which can be 0 for normal and 1 for fault) of whether a fault occurred in the corresponding time period. The system performs fault correlation analysis on the water pressure parameters, water level parameters, and current parameters respectively based on the historical fault record data, to measure the statistical correlation between each parameter and the occurrence of water pump faults. Specifically, Pearson correlation coefficients can be used to calculate the correlation between each monitoring parameter sequence and the fault label sequence. The system selects the monitoring index with the highest fault correlation degree according to the analysis results, and sets it as the single monitoring index used for subsequent preliminary fault determination. For example, if the average correlation coefficient of the current parameter is the largest, the current is selected as the single monitoring index.

[0024] Further, based on historical water pump fault monitoring records, the fault correlation analysis on the water pressure parameters, water level parameters, and current parameters respectively includes:

[0025] randomly selecting first historical water pump fault monitoring record data, and using Pearson correlation coefficients to perform fault correlation analysis on the water pressure parameters, water level parameters, and current parameters respectively, to obtain a first water pressure correlation coefficient, a first water level correlation coefficient, and a first current correlation coefficient; performing traversal analysis to obtain several water pressure correlation coefficients, several water level correlation coefficients, and several current correlation coefficients, and calculating the mean values to obtain a water pressure correlation degree, a water level correlation degree, and a current correlation degree.

[0026] Further, the monitoring index with the largest correlation degree among the water pressure correlation degree, water level correlation degree and current correlation degree is selected as the single monitoring index.

[0027] A multi-dimensional monitoring data set containing the water pump operating state and corresponding fault labels is obtained from the historical database. Each record contains at least the time series data of water pressure, water level and current in a fixed time window (for example, 5 minutes), and the label information of whether a water pump fault occurred in the time window. The label can be marked in Boolean type, with 1 indicating a fault and 0 indicating normal operation.

[0028] The system randomly selects a set of historical water pump fault monitoring records as the first analysis sample (i.e., the first historical water pump fault monitoring record data), constructs the corresponding three groups of monitoring parameter sequences and fault label sequences. The Pearson correlation coefficients are calculated for the above-mentioned three groups of monitoring parameters and fault label sequences, obtaining the first water pressure correlation coefficient, the first water level correlation coefficient and the first current correlation coefficient. After the above operation is completed, the system performs iterative analysis on multiple sample groups, i.e., different monitoring record samples are repeatedly extracted from the historical database, and the same correlation analysis steps are performed to obtain a plurality of water pressure correlation coefficients, a plurality of water level correlation coefficients and a plurality of current correlation coefficients. The mean value of each set of correlation coefficients is calculated to obtain the final water pressure correlation degree, water level correlation degree and current correlation degree. The monitoring index with the largest correlation degree is selected from the water pressure correlation degree, water level correlation degree and current correlation degree as the single monitoring index.

[0029] The fault warning sub-model is used to predict and obtain the first water pump fault probability in the future time window according to the single monitoring data sequence.

[0030] The fault warning sub-model is a time series prediction model trained based on a long short-term memory network (LSTM). The fault warning sub-model can identify the time series anomaly patterns in the data through supervised learning with a large amount of historical monitoring data and fault labels, and can predict the probability of a fault occurring in the future. The single monitoring sequence is input into the fault warning sub-model, and the first water pump fault probability in the future time window is output.

[0031] Further, the fault warning sub-model is constructed, including:

[0032] Based on the historical water pump fault monitoring records, a sample single monitoring data sequence set is collected, and the fault proportion of the sample single monitoring data sequence in the historical time window is obtained, which is set as the sample first fault probability, obtaining a sample first fault probability set; the long short-term memory network is trained with the sample single monitoring data sequence set as input and the sample first fault probability set as supervision until convergence, obtaining the fault warning sub-model.

[0033] Based on historical water pump failure monitoring records, a large number of sample single monitoring data sequence sets are collected, and for each sample data sequence, the proportion of water pump failure in the corresponding historical time window is calculated, referred to as the sample first failure probability, thereby obtaining the sample first failure probability set. Specifically, the single monitoring data sequence refers to the sensor monitoring data collected by backtracking a fixed length at a certain time point, which reflects the single key parameter change trend of the water pump of the disinfection cabinet. The sample first failure probability is the ratio of the actual failure duration of the water pump to the total duration of the window in the future preset time window from the time point as the failure probability label. By constructing a set of training samples with single monitoring data sequence as input and corresponding failure probability as output, using long short-term memory network (LSTM) as the model architecture, the network can effectively capture the potential time sequence correlation and nonlinear features in the time sequence. In the training process, the sample first failure probability set is used as the supervision label, and the loss function such as mean square error is used to optimize and iterate the network parameters until the prediction error of the model on the validation set converges and meets the preset performance indicators. Finally, the trained failure warning sub-model has the ability to quickly predict the failure probability of the input single monitoring data sequence, which can assist the system in realizing preliminary screening and warning of water pump failure in real-time monitoring, improving the timeliness and accuracy of failure warning.

[0034] If the first water pump failure probability is greater than the first failure probability threshold, the system will start the multi-dimensional monitoring data acquisition mechanism. Specifically, according to the multi-dimensional monitoring indicators, the system synchronously acquires the multi-dimensional monitoring data sequence of the disinfection cabinet in the preset time period through multiple sensors. These monitoring indicators at least include key operating parameters such as water pressure parameters, water level parameters, and current parameters. The multi-dimensional monitoring data sequence covers the entire preset time period and can comprehensively reflect the operating state and environmental changes of the water pump of the disinfection cabinet, providing rich time sequence information for subsequent failure diagnosis and risk assessment.

[0035] When the first water pump failure probability predicted by the failure warning sub-model exceeds the preset first failure probability threshold, the system will start the multi-dimensional monitoring data acquisition mechanism. Specifically, according to the multi-dimensional monitoring indicators, the system synchronously acquires the multi-dimensional monitoring data sequence of the disinfection cabinet in the preset time period through multiple sensors. These monitoring indicators at least include key operating parameters such as water pressure parameters, water level parameters, and current parameters. The multi-dimensional monitoring data sequence covers the entire preset time period and can comprehensively reflect the operating state and environmental changes of the water pump of the disinfection cabinet, providing rich time sequence information for subsequent failure diagnosis and risk assessment.

[0036] Using the failure warning parent model, the second water pump failure probability in the future time window is predicted and obtained according to the multi-dimensional monitoring data sequence.

[0037] The fault early warning parent model is based on a long short-term memory network (LSTM) structure, and can effectively capture the complex time sequence relationship and potential fault characteristics between the multi-dimensional monitoring indicators through training on a large amount of historical multi-dimensional monitoring data and corresponding fault conditions. The multi-dimensional monitoring data sequence is input into the fault early warning parent model, and the model is calculated through multiple layers of recurrent neural network to output a scalar value representing the second fault probability of the water pump in the future time window.

[0038] Further, the fault early warning parent model is constructed, including:

[0039] Based on the historical water pump fault monitoring records, a sample multi-dimensional monitoring data sequence set is collected, and the fault proportion of the sample multi-dimensional monitoring data sequence in the historical time window is obtained, which is set as the sample second fault probability, to obtain a sample second fault probability set. The long short-term memory network is trained to convergence using the sample multi-dimensional monitoring data sequence set and the sample second fault probability set, to obtain the fault early warning parent model.

[0040] Based on the historical water pump fault monitoring records, a large number of sample multi-dimensional monitoring data sequence sets are collected, each sample data sequence containing multiple key monitoring indicators such as water pressure parameters, water level parameters and current parameters. These multi-dimensional data sequences reflect the comprehensive operation state of the water pump of the disinfection cabinet. In the corresponding historical time window, the ratio of the duration of the water pump fault to the total duration of the window is calculated and defined as the second fault probability of the sample, thereby obtaining a set of sample second fault probability sets. Subsequently, the collected sample multi-dimensional monitoring data sequence set is used as input, and the sample second fault probability set is used as a supervision label to train the long short-term memory network (LSTM) model. The network parameters are repeatedly iterated and optimized until the prediction error of the model converges on the validation set, and finally a fault early warning parent model is obtained which can accurately predict the future fault probability of the water pump based on the multi-dimensional monitoring data sequence.

[0041] Further, the long short-term memory network is trained to convergence using the sample multi-dimensional monitoring data sequence set and the sample second fault probability set, to obtain the fault early warning parent model, including:

[0042] The sample multi-dimensional monitoring data sequence set and the sample second fault probability set are used as training data and equally divided into K parts. K times are selected with replacement in the K training sets, a first training set is constructed, K times are iteratively selected, and K training sets are obtained, wherein K is an integer greater than 10. The K training sets are used to supervise the training of the long short-term memory network, and K fault early warning branches meeting the convergence condition are obtained to construct the fault early warning parent model.

[0043] Specifically, a sample multi-dimensional monitoring data sequence set and a sample second failure probability set are taken as training data; the training data is equally divided into K parts, where K is an integer greater than 10; a random sampling method with replacement is used to randomly select K samples from the K parts of data K times to construct a first training set, and the same way is iteratively repeated K times to obtain K training sets; for each training set, a long short-term memory network (LSTM) model is trained respectively, which can process time series input data and learn the time series dependence relationship therein. During the training process, each LSTM model takes the multi-dimensional monitoring data sequence in the subset as input, and takes the corresponding second failure probability as the supervision label, and uses the gradient descent method to optimize the network parameters until the loss function value of the model on the validation set meets the convergence condition. When all K LSTM models are trained and meet the preset performance standard, the K trained and converged models are taken as K failure warning branches to jointly construct the final failure warning parent model.

[0044] Further, using the failure warning parent model, a second water pump failure probability in a future time window is predicted and obtained according to the multi-dimensional monitoring data sequence, comprising:

[0045] If the first water pump failure probability is greater than the first failure probability threshold, a first failure probability overflow value is calculated; the branch selection number P is analyzed and obtained according to the first failure probability overflow value; P failure warning branches are randomly selected from the K failure warning branches of the failure warning parent model, P failure probabilities are predicted according to the multi-dimensional monitoring data sequence, and a second water pump failure probability is obtained after mean calculation.

[0046] When the failure warning parent model is used to predict the multi-dimensional monitoring data sequence to obtain the second water pump failure probability in the future time window, first, it is judged whether the first water pump failure probability output by the failure warning sub-model exceeds the preset first failure probability threshold; if it exceeds, the excess part is calculated and recorded as the first failure probability overflow value, which reflects the abnormality degree of the current water pump operating state compared with the primary warning threshold. Then, based on the proportional relationship between the first failure probability overflow value and the maximum failure probability overflow value in the historical record, a branch adjustment coefficient is analyzed and determined, and the adjustment coefficient is multiplied by the total branch number K of the failure warning parent model and then rounded to obtain the number P of failure warning branches that need to participate in the current prediction. Next, P sub-models are randomly selected from the K trained failure warning branches, and the multi-dimensional monitoring data sequence obtained at the current time is input into each sub-model for prediction, and P independent second water pump failure probability values are output. Finally, the arithmetic mean of the P failure probabilities is taken as the final output second water pump failure probability, which is used to determine whether to trigger the formal failure warning of the water pump.

[0047] Further, the branch selection quantity P is obtained according to the first failure probability overflow value, comprising:

[0048] The ratio of the first failure probability overflow value to the maximum failure probability overflow value in the historical monitoring record is set as a branch adjustment coefficient; and the branch adjustment coefficient is multiplied by K to obtain the branch selection quantity P.

[0049] Specifically, the first failure probability overflow value is subjected to ratio operation with the maximum failure probability overflow value in the historical monitoring record, and the ratio is set as a branch adjustment coefficient, which is used to reflect the severity of the current abnormal state in the historical sample; the branch adjustment coefficient is multiplied by the total branch quantity K in the trained failure warning parent model, and the product is rounded down to obtain the final branch selection quantity P. This method realizes that when the first-stage warning probability is low, only a small number of model branches are called to maintain the operation efficiency, and when the warning probability significantly exceeds the threshold, more branches are dynamically called in to enhance the prediction stability and result reliability, thereby effectively improving the response flexibility and precision adaptability of the overall failure alarm mechanism.

[0050] If the second water pump failure probability is greater than the second failure probability threshold, a water pump failure warning of the disinfection cabinet is performed, wherein the second failure probability threshold is greater than the first failure probability threshold.

[0051] If the second water pump failure probability is greater than the second failure probability threshold, it is considered that the possibility of water pump failure has reached a high risk level set by the system, and the system will trigger the water pump failure warning operation of the disinfection cabinet, and send an alarm signal to the user or the maintenance system in time to prompt that the current equipment may have operation abnormities and needs to be repaired or replaced.

[0052] The second failure probability threshold is a more stringent risk judgment standard relative to the first failure probability threshold, and when specifically set, the value of the second failure probability threshold is higher than that of the first failure probability threshold, forming two different precision level warning judgment thresholds of preliminary screening and review, thereby constructing a double warning system of “low threshold warning + high threshold verification”.

[0053] In summary, the embodiments of the present application have at least the following technical effects:

[0054] Firstly, a single monitoring data sequence of the disinfection cabinet in a preset time zone is obtained through sensor monitoring according to a single monitoring index. Then, a first water pump failure probability in a future time window is predicted according to the single monitoring data sequence by using a failure warning sub-model. If the first water pump failure probability is greater than a first failure probability threshold, a multi-dimensional monitoring data sequence of the disinfection cabinet in a preset time period is obtained through sensor monitoring according to multi-dimensional monitoring indexes. Then, a second water pump failure probability in the future time window is predicted according to the multi-dimensional monitoring data sequence by using a failure warning parent model. If the first water pump failure probability is greater than the first failure probability threshold, a first failure probability overflow value is calculated. A branch selection number P is obtained according to the first failure probability overflow value. P failure warning branches are randomly selected from K failure warning branches of the failure warning parent model, and P failure probabilities are predicted according to the multi-dimensional monitoring data sequence. The second water pump failure probability is obtained by averaging the P failure probabilities. If the second water pump failure probability is greater than a second failure probability threshold, a water pump failure warning of the disinfection cabinet is performed, wherein the second failure probability threshold is greater than the first failure probability threshold. The technical problem of insufficient accuracy of disinfection cabinet failure warning in the prior art is solved. By using the double warning mechanism combining single monitoring index preliminary screening and multi-dimensional monitoring index review, the technical effect of improving the accuracy of disinfection cabinet failure warning is achieved.

[0055] In the second embodiment, based on the same inventive concept as the intelligent disinfection cabinet failure warning method based on sensor monitoring in the foregoing embodiments, as shown in the accompanying drawings, the present application provides an intelligent disinfection cabinet failure warning device based on sensor monitoring, wherein the device comprises: Figure 2

[0056] The monitoring module 11 obtains a single monitoring data sequence of the disinfection cabinet in a preset time zone through sensor monitoring according to a single monitoring index. The first failure prediction module 12 predicts a first water pump failure probability in a future time window according to the single monitoring data sequence by using a failure warning sub-model. The judgment module 13 obtains a multi-dimensional monitoring data sequence of the disinfection cabinet in a preset time period through sensor monitoring according to multi-dimensional monitoring indexes if the first water pump failure probability is greater than a first failure probability threshold. The second failure prediction module 14 predicts a second water pump failure probability in the future time window according to the multi-dimensional monitoring data sequence by using a failure warning parent model. The warning module 15 performs a water pump failure warning of the disinfection cabinet if the second water pump failure probability is greater than a second failure probability threshold, wherein the second failure probability threshold is greater than the first failure probability threshold.

[0057] Further, the monitoring module 11 is used to perform the following method:

[0058] ​The multi-dimensional monitoring index for water pump fault analysis is acquired, wherein the multi-dimensional monitoring index at least includes a water pressure parameter, a water level parameter and a current parameter; based on historical water pump fault monitoring records, the water pressure parameter, the water level parameter and the current parameter are analyzed for fault correlation respectively, and a single monitoring index is set according to the analysis result.

[0059] Further, the monitoring module 11 is used to execute the following method:

[0060] A first historical water pump fault monitoring record data is randomly selected, and the water pressure parameter, the water level parameter and the current parameter are analyzed for fault correlation respectively by using the Pearson correlation coefficient to obtain a first water pressure correlation coefficient, a first water level correlation coefficient and a first current correlation coefficient; traversal analysis is performed to obtain a plurality of water pressure correlation coefficients, a plurality of water level correlation coefficients and a plurality of current correlation coefficients, and the water pressure correlation degree, the water level correlation degree and the current correlation degree are obtained after mean calculation.

[0061] Further, the monitoring module 11 is used to execute the following method:

[0062] The monitoring index with the largest correlation degree among the water pressure correlation degree, the water level correlation degree and the current correlation degree is selected as the single monitoring index.

[0063] Further, the first fault prediction module 12 is used to execute the following method:

[0064] Based on the historical water pump fault monitoring records, a sample single monitoring data sequence set is collected, and a fault proportion of the sample single monitoring data sequence in a historical time window is obtained as a sample first fault probability to obtain a sample first fault probability set; the sample single monitoring data sequence set is used as input, and the sample first fault probability set is used as supervision to train a long short-term memory network until convergence to obtain a fault early warning sub-model.

[0065] Further, the second fault prediction module 14 is used to execute the following method:

[0066] Based on the historical water pump fault monitoring records, a sample multi-dimensional monitoring data sequence set is collected, and a fault proportion of the sample multi-dimensional monitoring data sequence in a historical time window is obtained as a sample second fault probability to obtain a sample second fault probability set; the sample multi-dimensional monitoring data sequence set and the sample second fault probability set are used to train a long short-term memory network to convergence to obtain a fault early warning parent model.

[0067] Further, the second fault prediction module 14 is used to execute the following method:

[0068] The sample multidimensional monitoring data sequence set and the sample second failure probability set are taken as training data, and are equally divided into K parts, K times are selected with replacement in the K training sets, a first training set is constructed, K times are selected iteratively, and K training sets are obtained, wherein K is an integer greater than 10; the K training sets are used to supervise the training of the long short-term memory network, and K failure warning branches meeting the convergence condition are obtained, and a failure warning parent model is constructed.

[0069] Further, the second failure prediction module 14 is configured to perform the following method:

[0070] If the first water pump failure probability is greater than the first failure probability threshold, a first failure probability overflow value is calculated; a branch selection number P is obtained according to the first failure probability overflow value; P failure warning branches are randomly selected from the K failure warning branches of the failure warning parent model, P failure probabilities are predicted according to the multidimensional monitoring data sequence, and a second water pump failure probability is obtained after mean calculation.

[0071] Further, the second failure prediction module 14 is configured to perform the following method:

[0072] The ratio of the first failure probability overflow value to the historical maximum failure probability overflow value is taken as a branch adjustment coefficient; the branch adjustment coefficient is multiplied by K to obtain the branch selection number P.

[0073] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0074] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0075] The present application is only an exemplary description of the present application, and should be considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalents, the present application intends to include these modifications and changes.

Claims

1. A fault alarm method for intelligent disinfection cabinets based on sensor monitoring, characterized in that, The method includes: Based on a single monitoring indicator, a single monitoring data sequence of the disinfection cabinet within a preset time zone is obtained through sensor monitoring. Using the fault early warning sub-model, the probability of the first water pump failure within a future time window is predicted based on the single monitoring data sequence. If the failure probability of the first water pump is greater than the first failure probability threshold, the multidimensional monitoring data sequence of the disinfection cabinet within a preset time period is obtained through sensor monitoring according to the multidimensional monitoring indicators. Using the fault early warning parent model, the probability of second water pump failure within a future time window is predicted based on the multi-dimensional monitoring data sequence. If the failure probability of the first water pump is greater than the first failure probability threshold, the first failure probability overflow value is calculated, and the number of branches to be selected P is obtained based on the analysis of the first failure probability overflow value. Within the K fault warning branches of the fault warning parent model, P fault warning branches are randomly selected, and P fault probabilities are predicted based on the multidimensional monitoring data sequence. The average value is then used to calculate the second water pump fault probability. If the failure probability of the second water pump is greater than the second failure probability threshold, a water pump failure warning is issued for the disinfection cabinet, wherein the second failure probability threshold is greater than the first failure probability threshold. Configure a single monitoring indicator, including: Obtain multidimensional monitoring indicators for water pump failure analysis, wherein the multidimensional monitoring indicators include at least water pressure parameters, water level parameters, and current parameters; Based on historical water pump fault monitoring records, fault correlation analysis was performed on the water pressure parameters, water level parameters, and current parameters respectively, and a single monitoring index was set according to the analysis results. Based on historical water pump fault monitoring records, fault correlation analysis was performed on the water pressure parameters, water level parameters, and current parameters, including: Randomly select the first historical water pump fault monitoring record data, and use Pearson correlation coefficient to perform fault correlation analysis on the water pressure parameter, water level parameter, and current parameter respectively to obtain the first water pressure correlation coefficient, the first water level correlation coefficient, and the first current correlation coefficient; By performing a comprehensive analysis, several water pressure correlation coefficients, several water level correlation coefficients, and several current correlation coefficients are obtained. After calculating the average values, the water pressure correlation degree, water level correlation degree, and current correlation degree are obtained. Construct a fault early warning sub-model, including: Based on historical pump failure monitoring records, a sample single monitoring data sequence set is collected, and the failure proportion of the sample single monitoring data sequence in the historical time window is obtained, which is set as the first failure probability of the sample, and the first failure probability set of the sample is obtained. Using the single monitoring data sequence set of the sample as input and the first fault probability set of the sample as supervision, a long short-term memory network is trained until convergence to obtain a fault early warning sub-model. Construct a fault early warning parent model, including: Based on historical pump failure monitoring records, a sample multidimensional monitoring data sequence set is collected, and the failure proportion of the sample multidimensional monitoring data sequence in the historical time window is obtained, which is set as the sample second failure probability, and the sample second failure probability set is obtained. Using the sample multidimensional monitoring data sequence set and the sample second fault probability set, a long short-term memory network is trained until convergence to obtain the fault early warning parent model.

2. The intelligent disinfection cabinet fault alarm method based on sensor monitoring according to claim 1, characterized in that, The monitoring indicator with the highest correlation among the water pressure correlation, water level correlation, and current correlation is selected as the single monitoring indicator.

3. The fault alarm method for intelligent disinfection cabinet based on sensor monitoring according to claim 1, characterized in that, Using the sample multidimensional monitoring data sequence set and the sample second fault probability set, a long short-term memory network is trained until convergence to obtain a fault early warning parent model, including: The sample multidimensional monitoring data sequence set and the sample second fault probability set are used as training data and divided into K equal parts. The first training set is constructed by selecting the K training sets with replacement K times. The selection is iterated K times to obtain K training sets, where K is an integer greater than 10. Using the K training sets, supervised training is performed on the Long Short-Term Memory network to obtain K fault warning branches that meet the convergence condition, and a fault warning parent model is constructed.

4. The intelligent disinfection cabinet fault alarm method based on sensor monitoring according to claim 1, characterized in that, The number of branches selected, P, is obtained based on the analysis of the first fault probability overflow value, including: The ratio of the first fault probability overflow value to the historical maximum fault probability overflow value is set as the branch adjustment coefficient. The number of branches selected, P, is obtained by multiplying the branch adjustment coefficient by K and rounding it down.

5. A fault alarm device for an intelligent disinfection cabinet based on sensor monitoring, characterized in that, For implementing the sensor-based intelligent disinfection cabinet fault alarm method according to any one of claims 1-4, the device comprises: Monitoring module: Based on a single monitoring indicator, it acquires a single monitoring data sequence of the disinfection cabinet within a preset time zone through sensor monitoring; First fault prediction module: Using the fault early warning sub-model, the probability of the first water pump failure within a future time window is predicted based on the single monitoring data sequence. Judgment module: If the failure probability of the first water pump is greater than the first failure probability threshold, the multidimensional monitoring data sequence of the disinfection cabinet within a preset time period is obtained by sensor monitoring according to the multidimensional monitoring indicators. The second fault prediction module: uses the fault early warning parent model to predict the probability of second water pump failure within a future time window based on the multi-dimensional monitoring data sequence; Early warning module: If the failure probability of the second water pump is greater than the second failure probability threshold, an early warning of water pump failure for the disinfection cabinet is issued, wherein the second failure probability threshold is greater than the first failure probability threshold.

Images