Disinfection quality assessment method and system based on big data

Abstract

The invention discloses a disinfection quality assessment method and system based on big data. The content comprises data acquisition, data cleaning, disinfection effect assessment, disinfection operation assessment and disinfection quality comprehensive assessment. The invention relates to the technical field of public health management, in particular to a disinfection quality assessment method and system based on big data, according to the scheme, a double-branch time sequence convolutional network and a graph attention mechanism are utilized to extract high-order spatial-temporal features from environment variables and biochemical indexes, and the disinfection quality assessment method and system based on the big data are obtained through cross-modal cross attention fusion in combination with context fine tuning. Outputting an accurate disinfection quality score; then adopting Monte Carlo simulation large-scale sampling in a historical disinfection operation parameter range, and combining Sobol sensitivity analysis to quantify contribution of disinfection operation parameters to disinfection quality score fluctuation; and finally, through reinforcement learning incremental iterative optimization in a discretized parameter space, providing an adaptive optimization strategy and a visual parameter contribution degree, and significantly improving coverage depth, accuracy and interpretability of evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of public health management, in particular to a disinfection quality evaluation method and system based on big data. BACKGROUND

[0002] Under the condition of gradually increasing health and safety requirements in the fields of medical health, food processing and public place management, traditional disinfection quality evaluation usually relies on manual sampling or fixed-point monitoring, obtains limited static or low-frequency time series data through temperature and humidity sensing, surface sampling and colony counting, and then makes simple determination combined with experience threshold; this method often faces problems such as limited coverage, long sampling period, single data dimension and difficulty in capturing dynamic correlation in actual application, and it is difficult to meet the needs of environmental monitoring and multi-dimensional index evaluation; at the same time, the setting of disinfection operation parameters is often based on experience or fixed standard, lacking quantitative analysis of the contribution of each key factor, and it is also difficult to quickly respond to new scene changes and standard updates, which restricts the evaluation accuracy and sustainable improvement capability. The development of big data technology brings possibilities for disinfection quality evaluation, which can integrate multi-channel information, build different scene evaluation models through mining and analysis, consider multiple influencing factors and their relationships, and update shared data in real time, so as to help regulatory departments and enterprise management departments to timely grasp the situation, find problems and improve, in order to meet the health and safety needs. SUMMARY

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present application provides a disinfection quality evaluation method and system based on big data, which is aimed at the problem that traditional disinfection effect evaluation relies on single time series or static features and is difficult to consider environmental dynamics and biochemical correlation; this scheme extracts high-order dynamics and correlation features of environmental variables and chemical microbial indicators through double-branch time series convolution network and graph attention mechanism respectively, and deeply fuses spatio-temporal information in cross-modal cross-attention, finally realizes disinfection quality score prediction combined with mapping output and context fine-tuning; for the problem that traditional disinfection evaluation relies on experience index, has limited coverage and is difficult to quantify the real influence of operation parameters, this scheme constructs a comprehensive parameter score mapping in the three-dimensional space of spraying rate, time length and drug concentration through Monte Carlo scene simulation, accurately identifies the contribution of parameters to disinfection quality score fluctuation combined with Sobol first-order sensitivity analysis, so as to focus on the most critical optimization direction; and through reinforcement learning strategy, it realizes real-time test and dynamic optimization in the discretized parameter set, and pushes disinfection evaluation to a new height of quantifiable, interpretable and self-adaptive.

[0004] The technical scheme adopted by the present application is as follows: a disinfection quality evaluation method based on big data, which comprises the following steps:

[0005] Step S1: Collecting data, collecting environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area;

[0006] Step S2: Cleaning data, hard checking the original field according to the preset threshold;

[0007] Step S3: Disinfection effect evaluation, extracting high-order dynamic features of environmental data and operation log data through double-branch time series convolution network respectively; at the same time, constructing time-sensitive feature map of chemical and microbial detection data by using graph attention network; then introducing cross-modal cross-attention mechanism, deep interaction between time series features and graph structure features is completed, high-dimensional fusion between space, time and variables is completed; finally outputting disinfection quality score;

[0008] Step S4: Disinfection operation evaluation, generating disinfection score distribution of parameter combination in the range of historical operation log data through Monte Carlo simulation, then quantifying the contribution of parameters to score fluctuation by using Sobol first-order sensitivity analysis, finally updating the score estimate value in the discretized parameter space by using reinforcement learning strategy, repeatedly testing and selecting the optimal disinfection operation parameter combination;

[0009] Step S5: Disinfection quality comprehensive evaluation, real-time collecting environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area to be evaluated; first, disinfection effect evaluation is performed to generate disinfection quality score; then, disinfection operation evaluation is performed to generate optimal disinfection operation parameter combination.

[0010] Further, in step S1, the data is collected, the environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area are collected; the environmental data includes temperature, humidity, air flow rate; the chemical and microbial detection data includes colony count, ATP fluorescence value, residual chemical concentration; the operation log data includes spraying time, spraying rate, drug concentration; the maintenance record data includes disinfection scheme version and period.

[0011] Further, in step S2, the data is cleaned, specifically, the original field is hard checked according to the preset threshold, and the sample points that do not meet the checking rule are directly excluded.

[0012] Further, in step S3, the disinfection effect evaluation specifically includes the following steps:

[0013] Step S31: Time series feature extraction, the environmental sequence composed of environmental data and the log sequence composed of operation log data are extracted by double-branch time series convolution network respectively to extract high-order dynamic features;

[0014] Step S32: graph attention learning, constructing graph nodes with chemical and microbial detection data, connecting nodes at the same time through dynamic adjacency strategy, using time dynamic adjacency matrix, and updating graph structure online according to time similarity and numerical correlation;

[0015] Step S33: spatio-temporal feature cross fusion, introducing bidirectional cross attention, inputting the time sequence feature sequence obtained in S31 and the graph feature obtained in S32 into the cross-domain attention module, performing deep interaction of environment, log and biochemical data, and generating fusion features;

[0016] Step S34: mapping output, predicting the disinfection quality score based on the fusion features, and making context fine-tuning according to the maintenance record data.

[0017] Further, in step S4, the disinfection operation evaluation specifically includes the following steps:

[0018] Step S41: Monte Carlo scenario simulation, randomly sampling parameter combinations in the value range of the disinfection operation parameters determined in the historical operation log according to uniform distribution, and inputting the parameters into the trained disinfection quality score model to obtain the corresponding score output;

[0019] Step S42: parameter sensitivity analysis, quantifying the contribution of operation parameters: spraying rate, spraying time, and drug concentration to the total variance of the final disinfection quality score;

[0020] Step S43: obtaining optimal disinfection operation parameters, specifically including the following steps:

[0021] Step S431: parameter combination space initialization, first constructing an action space according to historical disinfection records, each action representing a set of operation parameter combinations; then extracting boundaries from historical data to build a uniformly discretized action set;

[0022] Step S432: action score estimation and update, initializing score estimation for each action; then calling the disinfection quality evaluation model for the action to obtain the actual score; and updating the score estimation value of the action using the incremental method;

[0023] Step S433: optimal parameter output, outputting the current optimal disinfection parameter as a recommendation result.

[0024] Further, in step S5, the disinfection quality comprehensive evaluation is performed, specifically, the environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area are collected in real time; the disinfection effect is evaluated first to obtain a disinfection quality score, and the higher the score, the better the disinfection quality; then the disinfection operation is evaluated, and the optimal disinfection operation parameter combination is output, if the current disinfection parameter is not equal to the optimal disinfection parameter, the current disinfection operation is evaluated as the to-be-optimized disinfection operation; otherwise, the current disinfection operation is evaluated as the optimal disinfection operation.

[0025] The disinfection quality evaluation system based on big data provided by the application comprises a data collection module, a data cleaning module, a disinfection effect evaluation module, a disinfection operation evaluation module and a disinfection quality comprehensive evaluation module.

[0026] The data collection module collects the environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection site, and sends the data to the data cleaning module.

[0027] The data cleaning module receives the data sent by the data collection module, performs hard check on the original fields according to a preset threshold, and sends the data to the disinfection effect evaluation module, the disinfection operation evaluation module and the disinfection quality comprehensive evaluation module.

[0028] The disinfection effect evaluation module receives the data sent by the data cleaning module, extracts high-order dynamic features of the environmental data and operation log through a double-branch time sequence convolution network; at the same time, a graph attention network is used to construct a time-sensitive feature map of the chemical and microbial detection data; then a cross-modal cross-attention mechanism is introduced to perform deep interaction between the time sequence features and the graph structure features; finally, a disinfection quality score is output, and the data is sent to the disinfection operation evaluation module and the disinfection quality comprehensive evaluation module.

[0029] The disinfection operation evaluation module receives the data sent by the data cleaning module and the disinfection effect evaluation module, generates a disinfection score distribution of the parameter combination in the range of the historical operation log data through Monte Carlo simulation, then quantifies the contribution of the parameters to the score fluctuation by using Sobol first-order sensitivity analysis, finally incrementally updates the score estimate value in the discretized parameter space by using a reinforcement learning strategy, repeatedly tests and selects the optimal disinfection operation parameter combination, and sends the data to the disinfection quality comprehensive evaluation module.

[0030] The disinfection quality comprehensive evaluation module receives the data sent by the data cleaning module, the disinfection effect evaluation module and the disinfection operation evaluation module, collects the environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area in real time; first, the disinfection effect is evaluated to obtain a disinfection quality score, and the higher the score, the better the disinfection quality; then, the disinfection operation is evaluated, and the optimal disinfection operation parameter combination is output.

[0031] The application has the following beneficial effects by adopting the above scheme:

[0032] (1) In view of the problem that traditional disinfection effect evaluation relies on single time sequence or static features and is difficult to take into account environmental dynamics and biochemical correlations, the scheme extracts high-order dynamics and correlation features of environmental variables and chemical microbial indicators through a double-branch time sequence convolution network and a graph attention mechanism, deeply fuses time and space information in cross-modal cross-attention, and finally realizes disinfection quality score prediction in combination with mapping output and context fine-tuning; both the time sequence perception of the environment and operation logs are strengthened, and the structural dependence between colony count, ATP fluorescence value and chemical residues is captured, which significantly improves the evaluation accuracy and generalization ability, and enhances the model transparency and operability.

[0033] (2) In view of the problem that traditional disinfection evaluation relies on empirical indicators, has limited coverage and is difficult to quantify the real influence of operation parameters, the scheme constructs a comprehensive parameter score mapping in the three-dimensional space of spraying rate, time length and drug concentration through Monte Carlo scene simulation, accurately identifies the contribution of parameters to disinfection quality score fluctuations in combination with Sobol first-order sensitivity analysis, so as to focus on the most critical optimization direction; and realizes real-time testing and dynamic optimization through incremental iteration in the discretized parameter set through reinforcement learning strategy; not only significantly improves the depth and accuracy of evaluation, but also provides quantitative and visual parameter contribution, strengthens the transparency and decision support ability of evaluation, and finally pushes disinfection evaluation to a new height of quantifiable, interpretable and self-adaptive. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 A schematic diagram of the disinfection quality evaluation method based on big data provided by the application;

[0035] Figure 2 A schematic diagram of the disinfection quality evaluation system based on big data provided by the application;

[0036] Figure 3 A schematic diagram of step S3;

[0037] Figure 4 A schematic diagram of step S4.

[0038] The accompanying drawings are used to provide a further understanding of the application, and constitute a part of the specification, together with the embodiments of the application, to explain the application, and do not constitute a limitation on the application. DETAILED DESCRIPTION

[0039] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments; based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0040] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore cannot be understood as a limitation on the present application.

[0041] Embodiment one, refer to Figure 1 The present application provides a disinfection quality evaluation method based on big data, which comprises the following steps:

[0042] Step S1: collecting data, collecting environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area; the environmental data includes temperature, humidity, air flow rate; the chemical and microbial detection data includes colony count, ATP fluorescence value, residual chemical concentration; the operation log data includes spraying time, spraying rate, drug concentration; the maintenance record data includes disinfection scheme version and cycle;

[0043] Step S2: cleaning data, hard checking the original field according to the preset threshold value;

[0044] Step S3: disinfection effect evaluation, extracting high-order dynamic features of environmental data and operation log data through double-branch time sequence convolution network; at the same time, constructing time-sensitive feature map of chemical and microbial detection data by using graph attention network; then introducing cross-modal cross-attention mechanism, deep interaction between time sequence features and graph structure features is carried out, high-dimensional fusion between space, time and variables is completed; finally, outputting disinfection quality score;

[0045] Step S4: disinfection operation evaluation, generating disinfection score distribution of parameter combination in the range of historical operation log data through Monte Carlo simulation, then quantifying the contribution of parameters to score fluctuation by using Sobol first-order sensitivity analysis, finally updating the score estimate value in the discretized parameter space by using reinforcement learning strategy, repeatedly testing and selecting the optimal disinfection operation parameter combination;

[0046] Step S5: comprehensive evaluation of disinfection quality, real-time collection of environmental data, chemical and microbial detection data, operation log data and maintenance record data of the disinfection area to be evaluated; first, disinfection effect evaluation is performed to generate a disinfection quality score; then, disinfection operation evaluation is performed to generate an optimal disinfection operation parameter combination.

[0047] Embodiment Two, see Figure 1 This embodiment is based on the above embodiment, in step S2, the cleaning data, specifically, the original field is hard checked according to the preset threshold; temperature sequence: allowed range ; humidity sequence: allowed range ; air flow rate sequence: allowed range ; colony count sequence: allowed range ; ATP fluorescence value sequence: allowed range ; residual chemical concentration sequence: allowed range ; spraying rate sequence: allowed range ; spraying duration sequence: allowed range ; medicament concentration sequence: allowed range ; samples that do not meet any of the above conditions are directly excluded.

[0048] Embodiment Three, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, in step S3, the disinfection effect evaluation, specifically includes the following steps:

[0049] Step S31: time sequence feature extraction, the three-dimensional environmental sequence of temperature , humidity , and air flow rate , and the three-dimensional log sequence of spraying duration , spraying rate , and medicament concentration are extracted through a double-branch time sequence convolution network to obtain high-order dynamic features, which are represented as follows:

[0050] ;

[0051] wherein, represents the convolution feature vector of the environmental branch at time t, represents a linear rectification function, n represents the index of the convolution kernel of the environmental branch, and N represents the total number of convolution kernels of the environmental branch, and respectively represent the learnable parameter matrix and the bias vector of the nth convolution kernel of the environmental branch, represents the set of learnable dilation steps of the environmental branch, represents the input of the environmental branch at time , wherein Indicates the transpose symbol. , and They represent the times at time 1 and 2 respectively. Temperature, humidity, and air velocity; Let represent the convolutional feature vector of the log branch at time t, m represent the index of the convolutional kernel of the log branch, and M represent the total number of convolutional kernels of the log branch. and Let represent the learnable parameter matrix and bias vector of the m-th convolutional kernel in the log branch, respectively. This represents the set of learnable expansion steps for log branches. Indicates at time Log branch input, , and They represent the times respectively. Spraying duration, spraying rate, and pesticide concentration;

[0052] Step S32: Graph attention learning, colony counting ATP fluorescence value Residual chemical concentration The graph nodes are constructed by connecting nodes that occur at the same time using a dynamic adjacency strategy, employing a temporal dynamic adjacency matrix. The graph structure is updated online based on temporal similarity and numerical correlation, as shown below:

[0053] ;

[0054] Where i, j, and k represent the node indices, and the chemical and microbiological detection node features. , Let represent the adjacency matrix at time t, and the set of neighboring nodes of the i-th node. By dynamic adjacency matrix The specific determination process is as follows: First, the characteristic values ​​of the three nodes—colon count, ATP fluorescence value, and chemical residue concentration—at the same time t are standardized. Then, the Pearson correlation coefficient is calculated pairwise to obtain a 3×3 original similarity matrix. Next, the similarity value of each row is divided by the sum of the values ​​in that row to complete row normalization. Subsequently, for each row of normalized data, the two positions with the largest values ​​are retained, and their corresponding matrix entries are set to 1, while the remaining positions are set to 0. Finally, the matrix... The column indices of all elements in the i-th row that are 1 form the neighbor set of the i-th node. ; This represents the original attention scores of nodes i and j at time t. This represents a linear rectified function with leakage. denote the learnable attention mechanism parameters, denote the concatenation operation, denote the shared learnable linear transformation matrix, and denote the chemical and microbiological detection feature values of the i-th and j-th node at time t, denote the normalized attention weights, denote the exponential function with a natural constant as the base, denote the original attention scores of nodes i and k at time t, denote the updated graph feature vector of node i;

[0055] Step S33: Spatio-temporal feature cross-fusion, introduce bidirectional cross-attention, input the time sequence feature sequence obtained in S31 and the graph feature obtained in S32 into the cross-domain attention module for deep interaction of environment, log and biochemical data, denoted as follows:

[0056] ;

[0057] wherein, , and denote the query, key and value vectors, , , and denote the learnable mapping matrix, denote the scaling factor, taking a value in the range [0.1, 10]; denote the output features after fusion, denote the normalized exponential function;

[0058] Step S34: Mapping output, based on the fusion features , predict the disinfection quality score, and make context fine-tuning according to the historical scheme version and the period , denoted as follows:

[0059] ;

[0060] wherein, denote the version number of the historical disinfection scheme, denote the specific number of days of the historical disinfection period, denote the context fine-tuning vector, , and denote the learnable context encoding parameters, denote the features after context concatenation, represents a disinfection quality score, the higher the score, the better the disinfection quality; and respectively represent the learnable mapping output weight and bias.

[0061] By performing the above operation, in view of the problem that the traditional disinfection effect evaluation relies on a single time sequence or static feature and is difficult to take into account the environmental dynamics and biochemical correlation, the present scheme extracts high-order dynamic and correlation features of environmental variables and chemical microbial indicators through a double-branch time sequence convolution network and a graph attention mechanism, and deeply fuses the space-time information in the cross-modal cross-attention, and finally realizes the disinfection quality score prediction combined with the mapping output and the context fine-tuning; both the time sequence perception of the environment and the operation log are strengthened, and the structural dependence between the colony count, the ATP fluorescence value and the chemical residue is captured, which significantly improves the evaluation accuracy and the generalization ability, and enhances the model transparency and operability.

[0062] Embodiment four, refer to Figure 1 and Figure 4 , this embodiment is based on the above-mentioned embodiment, in step S4, the disinfection operation evaluation specifically includes the following steps:

[0063] Step S41: Monte Carlo scenario simulation, through the method of Monte Carlo simulation, G groups of parameter combinations are randomly sampled in the value interval of the disinfection operation parameters determined in the historical operation log according to uniform distribution, and each group of parameters is input into the trained disinfection quality score model to obtain the corresponding score output, which is represented as follows:

[0064] ;

[0065] Among them, represents the disinfection quality score model constructed in step S3, g represents the index of the Monte Carlo sampling number, G represents the total number of Monte Carlo sampling, and the value range is , , and respectively represent the spraying rate, the spraying time and the drug concentration of the gth sampling, represents the score output of the gth sampling parameter combination by the model, and respectively represent the lower limit and the upper limit of the spraying rate, and respectively represent the lower limit and the upper limit of the spraying time, and respectively represent the lower limit and the upper limit of the drug concentration, and all the above lower limits and upper limits respectively correspond to the minimum value and the maximum value in all the operation log data remaining after step S2 is performed; represents the uniform distribution on the interval , denotes from a uniform distribution , sampled, and by the same token;

[0066] Step S42: parameter sensitivity analysis, first calculate the total variance of the score of all sampling parameters; then fix the value of each parameter including spray rate, spray duration, and drug concentration, calculate the conditional expectation of the score, and take the variance of the conditional expectation, then take the ratio of the total variance as the first-order sensitivity index of the parameter, quantify the contribution of the operation parameter to the total variance of the final disinfection quality score, denoted as follows:

[0067] ;

[0068] wherein, denotes the score random variable obtained by Monte Carlo simulation, denotes a single parameter to be analyzed, denotes the variance, is the spray rate, is the spray duration, is the drug concentration; denotes the total variance of the score, denotes the conditional expectation variance of Y after fixing the parameter ; denotes the Sobol first-order sensitivity index of the parameter , indicating the contribution of the operation parameter to the disinfection quality score, the higher the value of , the greater the influence of the operation parameter on the disinfection quality score;

[0069] Step S43: obtaining optimal disinfection operation parameters, specifically including the following steps:

[0070] Step S431: parameter combination space initialization, first construct the action space Act according to the historical disinfection records, wherein each action denotes a set of operation parameter combinations: ; then extract the boundary from the historical data to build a uniformly discretized action set, denoted as follows:

[0071] ;

[0072] wherein, denotes the spray rate discrete set, denotes the spray rate discrete step, denotes the standard deviation of the spray rate of all data samples; denotes the spray duration discrete set, denotes a spray duration discrete step, denotes a spray duration standard deviation of all data samples; denotes a medicament concentration discrete set, denotes a medicament concentration discrete step, denotes a medicament concentration standard deviation of all data samples;

[0073] Step S432: Action score estimation and update, for each action Initialize the score estimation, denoted as follows:

[0074] ;

[0075] wherein, denotes an initial estimation value of the evaluation, and is specifically taken as an average value of the disinfection quality score of all sample data; denotes the action that has been tried for the ith time; then, at the ith parameter combination update, the action with the highest score estimation value is selected from the action set: wherein, denotes the score estimation value of the action after the ith parameter combination update; then, for the action , the disinfection quality evaluation model is called to obtain the actual score ; and the score estimation value of the action is updated using the incremental method, denoted as follows:

[0076] ;

[0077] wherein, denotes the score estimation value of the action after the ith parameter combination update; denotes the score estimation value of the action after the (i-1)th parameter combination update, denotes a disinfection optimization learning rate, denotes the number of times the action has been tried before the ith parameter combination update, denotes the number of times the action has been tried before the (i-1)th parameter combination update;

[0078] Step S433: Optimal parameter output, outputting the current optimal disinfection parameter as a recommended result, denoted as follows:

[0079] ;

[0080] wherein, denotes the optimal disinfection operation parameter combination.​

[0081] By performing the above operation, in view of the problems that the traditional disinfection evaluation relies on empirical indicators, has limited coverage and is difficult to quantify the real impact of operation parameters, the scheme constructs a comprehensive parameter score mapping in the three-dimensional space of spraying rate, duration and drug concentration through Monte Carlo scene simulation, accurately identifies the contribution of parameters to the fluctuation of disinfection quality score through Sobol first-order sensitivity analysis, so as to focus on the most critical optimization direction; and through reinforcement learning strategy, incrementally iterates in the discretized parameter set, realizes real-time test and dynamic optimization; not only significantly improves the coverage depth and accuracy of the evaluation, but also provides quantitative and visual parameter contribution, strengthens the transparency and decision support capability of the evaluation, and finally pushes the disinfection evaluation to a new height of quantifiable, interpretable and self-adaptive.

[0082] Embodiment five, see Figure 1 The embodiment is based on the above-mentioned embodiment, in step S5, the disinfection quality comprehensive evaluation, specifically real-time collection of environmental data, chemical and microbial detection data, operation log data and maintenance record data in the disinfection area; first, the disinfection effect evaluation is carried out, and the disinfection quality score is obtained, the higher the score, the better the disinfection quality; then, the disinfection operation evaluation is carried out, and the optimal disinfection operation parameter combination is output, if the current disinfection parameter is not equal to the optimal disinfection parameter, the current disinfection operation is evaluated as the optimized disinfection operation; otherwise, the current disinfection operation is evaluated as the optimal disinfection operation.

[0083] Embodiment six, see Figure 1 And Figure 2 Figure 2 The embodiment is based on the above-mentioned embodiment, the disinfection quality evaluation system based on big data provided by the application comprises a data collection module, a data cleaning module, a disinfection effect evaluation module, a disinfection operation evaluation module and a disinfection quality comprehensive evaluation module.

[0084] The data collection module collects environmental data, chemical and microbial detection data, operation log data and maintenance record data in the disinfection place, and sends the data to the data cleaning module;

[0085] The data cleaning module receives the data sent by the data collection module, performs hard check on the original field according to the preset threshold, and sends the data to the disinfection effect evaluation module, the disinfection operation evaluation module and the disinfection quality comprehensive evaluation module;

[0086] The disinfection effect evaluation module receives the data sent by the cleaning data module, extracts high-order dynamic characteristics of the environmental data and the operation log through a double-branch time sequence convolution network; simultaneously, constructs a time-sensitive feature map of the chemical and microbial detection data by using a graph attention network; then, introduces a cross-modal cross-attention mechanism to perform deep interaction between the time sequence characteristics and the graph structure characteristics, and completes high-dimensional fusion between space, time and variables; finally, outputs a disinfection quality score, and sends the data to the disinfection operation evaluation module and the disinfection quality comprehensive evaluation module.

[0087] The disinfection operation evaluation module receives the data sent by the cleaning data module and the disinfection effect evaluation module, generates a disinfection score distribution of parameter combinations in the range of historical operation log data through Monte Carlo simulation, then quantifies the contribution of parameters to score fluctuation by using Sobol first-order sensitivity analysis, finally performs incremental update on the score estimated value in the discretized parameter space by using a reinforcement learning strategy, repeatedly tests and selects an optimal disinfection operation parameter combination, and sends the data to the disinfection quality comprehensive evaluation module.

[0088] The disinfection quality comprehensive evaluation module receives the data sent by the cleaning data module, the disinfection effect evaluation module and the disinfection operation evaluation module, collects environmental data, chemical and microbial detection data, operation log data and maintenance record data of a disinfection area in real time; first performs disinfection effect evaluation to obtain a disinfection quality score, and the higher the score is, the better the disinfection quality is; then performs disinfection operation evaluation to output an optimal disinfection operation parameter combination, if the current disinfection parameter is not equal to the optimal disinfection parameter, the current disinfection operation is evaluated as a to-be-optimized disinfection operation; otherwise, the current disinfection operation is evaluated as an optimal disinfection operation.

[0089] It should be noted that, in this document, the terms such as first and second are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or equipment.

[0090] Although the embodiments of the present application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and spirits of the present application.

[0091] The above describes the present application and its embodiments, which are not limited, and the drawings only show one of the embodiments of the present application, and the actual structure is not limited thereto. In general, if a person skilled in the art is inspired thereby, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution are not creative, and should belong to the protection scope of the present application.

Claims

1. A disinfection quality assessment method based on big data, characterized in that, The method includes the following steps: Step S1: Collect data, including environmental data, chemical and microbiological test data, operation log data, and maintenance record data of the disinfection area; Step S2: Clean the data and perform hard validation on the original fields according to preset thresholds; Step S3: Disinfection effect evaluation. High-order dynamic features of environmental data and operation logs are extracted by a dual-branch temporal convolutional network. At the same time, a graph attention network is used to construct a time-sensitive feature map of chemical and microbiological detection data. Then, a cross-modal cross-attention mechanism is introduced to deeply interact with temporal features and graph structure features to complete the high-dimensional fusion between space, time and variables. Finally, a disinfection quality score is output. Step S4: Disinfection operation evaluation. A disinfection score distribution with parameter combinations is generated within the historical operation log data range through Monte Carlo simulation. Then, the contribution of parameters to score fluctuations is quantified using Sobol first-order sensitivity analysis. Finally, the score estimate is incrementally updated in the discretized parameter space using a reinforcement learning strategy. The optimal combination of disinfection operation parameters is selected through repeated trials. Step S5: Comprehensive assessment of disinfection quality. Real-time collection of environmental data, chemical and microbiological test data, operation log data, and maintenance record data of the area to be assessed for disinfection. First, the disinfection effect is assessed to generate a disinfection quality score. Then, the disinfection operation is assessed to generate the optimal combination of disinfection operation parameters.

2. The disinfection quality assessment method based on big data according to claim 1, characterized in that: In step S3, the evaluation of the disinfection effect specifically includes the following steps: Step S31: Temporal feature extraction. The environmental sequence composed of environmental data and the log sequence composed of operation log data are used to extract high-order dynamic features through a dual-branch temporal convolutional network. Step S32: Graph attention learning, constructing graph nodes from chemical and microbiological detection data, connecting nodes at the same time through a dynamic adjacency strategy, using a temporal dynamic adjacency matrix, and updating the graph structure online based on temporal similarity and numerical correlation; Step S33: Spatiotemporal feature cross-fusion, introducing bidirectional cross-attention, inputting the temporal feature sequence obtained in S31 and the graph feature obtained in S32 into the cross-domain attention module to perform deep interaction of environmental, log and biochemical data to generate fused features; Step S34: Map the output, predict the disinfection quality score based on the fusion features, and fine-tune the context based on the maintenance record data.

3. The disinfection quality assessment method based on big data according to claim 1, characterized in that: In step S4, the disinfection operation evaluation specifically includes the following steps: Step S41: Monte Carlo scenario simulation. Using the Monte Carlo simulation method, within the value range of disinfection operation parameters determined in the historical operation log, parameter combinations are randomly sampled in a uniform distribution and input into the trained disinfection quality scoring model to obtain the corresponding score output. Step S42: Parameter sensitivity analysis, quantifying the contribution of operating parameters: spraying rate, spraying duration, and agent concentration to the total variance of the final disinfection quality score; Step S43: Obtain the optimal disinfection operation parameters, which specifically includes the following steps: Step S431: Initialize the parameter combination space. First, construct the action space based on historical disinfection records. Each action represents a set of operation parameter combinations. Then, extract the boundaries from the historical data and construct a uniformly discretized action set. Step S432: Action score estimation and update. Initialize the score estimate for each action; then call the disinfection quality assessment model for the action to obtain the actual score; then use the incremental method to update the score estimate of the action. Step S433: Optimal parameter output, output the current optimal disinfection parameters as the recommended result.

4. The disinfection quality assessment method based on big data according to claim 1, characterized in that: In step S1, the data collection includes environmental data, chemical and microbiological detection data, operation log data, and maintenance record data of the disinfection area. The environmental data includes temperature, humidity, and air velocity. The chemical and microbiological detection data includes colony count, ATP fluorescence value, and residual chemical concentration. The operation log data includes spraying duration, spraying rate, and agent concentration. The maintenance record data includes the disinfection plan version and cycle.

5. The disinfection quality assessment method based on big data according to claim 1, characterized in that: In step S2, the data cleaning specifically involves performing hard validation on the original fields according to a preset threshold, and directly removing sample points that do not meet the validation rules.

6. The disinfection quality assessment method based on big data according to claim 1, characterized in that: In step S5, the comprehensive evaluation of disinfection quality specifically involves real-time collection of environmental data, chemical and microbiological detection data, operation log data, and maintenance record data of the disinfection area; first, the disinfection effect is evaluated to obtain a disinfection quality score, with higher scores indicating better disinfection quality; Then, the disinfection operation is evaluated, and the optimal combination of disinfection operation parameters is output. If the current disinfection parameters are not equal to the optimal disinfection parameters, the current disinfection operation is evaluated as a disinfection operation to be optimized. Otherwise, the current disinfection operation will be evaluated as the optimal disinfection operation.

7. A disinfection quality assessment system based on big data, used to implement the disinfection quality assessment method based on big data as described in any one of claims 1-6, characterized in that: It includes a data collection module, a cleaning data module, a disinfection effect evaluation module, a disinfection operation evaluation module, and a comprehensive disinfection quality evaluation module.

8. The disinfection quality assessment system based on big data according to claim 7, characterized in that: The data acquisition module collects environmental data, chemical and microbiological test data, operation log data, and maintenance record data of the disinfection site, and sends the data to the cleaning data module. The cleaning data module receives data sent by the data acquisition module, performs hard validation on the original fields according to preset thresholds, and sends the data to the disinfection effect evaluation module, the disinfection operation evaluation module, and the comprehensive disinfection quality evaluation module. The disinfection effect evaluation module receives data sent by the cleaning data module and extracts high-order dynamic features from environmental data and operation logs through a dual-branch temporal convolutional network. At the same time, it uses a graph attention network to construct a time-sensitive feature map of chemical and microbial detection data. Subsequently, a cross-modal cross attention mechanism is introduced to deeply interact with temporal features and graph structure features. Finally, a disinfection quality score is output, and the data is sent to the disinfection operation assessment module and the comprehensive disinfection quality assessment module; The disinfection operation evaluation module receives data from the cleaning data module and the disinfection effect evaluation module. It generates a disinfection score distribution of parameter combinations within the historical operation log data range through Monte Carlo simulation. Then, it uses Sobol first-order sensitivity analysis to quantify the contribution of parameters to score fluctuations. Finally, it uses a reinforcement learning strategy to incrementally update the score estimate in the discretized parameter space. It repeatedly experiments and selects the optimal combination of disinfection operation parameters, and sends the data to the disinfection quality comprehensive evaluation module. The comprehensive disinfection quality assessment module receives data from the cleaning data module, the disinfection effect assessment module, and the disinfection operation assessment module. It collects environmental data, chemical and microbiological detection data, operation log data, and maintenance record data of the disinfection area in real time. First, it performs a disinfection effect assessment to obtain a disinfection quality score, with higher scores indicating better disinfection quality. Then, it performs a disinfection operation assessment to output the optimal combination of disinfection operation parameters.

Images