Method and system for testing suitability of non-invasive closed respiratory protection equipment
By building a fit test training sample library and using machine learning models, non-invasive fit testing of tight-fitting respiratory protective equipment can be achieved, which solves the problems of high cost and low efficiency in existing technologies, improves the accuracy and reliability of testing, and is suitable for actual production.
Patent Information
- Application Number
- CN202510583915.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-07
- Publication Date
- 2025-10-03
AI Technical Summary
The cost of suitability testing for existing tight-fitting respiratory protective equipment is high, and the testing process for users and occupational health departments is complicated, making it impossible to widely, timely and efficiently apply it to actual production.
A fit test training sample library is constructed, and a machine learning model is used to predict the fit level, including collecting head and facial data and tight-fitting respiratory protective equipment data. Through multiple model training and weighted calculations, non-invasive fit testing is achieved.
It significantly reduces inspection costs, improves efficiency, ensures the accuracy and reliability of inspections, and is suitable for widespread use in actual production to ensure the respiratory safety of workers.
Smart Images
Figure CN120744802A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of respiratory protection technology, and in particular to a method and system for testing the suitability of tight-fitting respiratory protection equipment. Background Art
[0002] In the field of respiratory protection technology for workers, respiratory protective equipment must be provided to ensure their work safety. Referring to relevant national standards, fit testing is an important basis for determining whether self-priming filter-type respiratory protective equipment is effective. However, the process of fit testing is very complicated. Specifically, fit testing requires professional technical equipment, conditions and personnel, and the inspection cost is high. Taking disposable masks as an example, it is generally difficult for its users and production units to conduct this inspection, and thus it cannot be widely used in actual production. At the same time, there are many types of tight-fitting respiratory protective equipment on the market, and wearers have no specific reference as a standard for selection when making a choice, resulting in the wearer being unable to correctly choose a tight-fitting respiratory protective equipment that is more suitable for themselves. For enterprises and governments, when distributing tight-fitting respiratory protective equipment to workers, they need to prepare a variety of candidates for them and adapt them to the workers' face shapes, which increases costs and is less efficient. Summary of the Invention
[0003] The present application provides a non-invasive, tight-fitting respiratory protective equipment fit testing method and system, aiming to solve the problems of high cost of fit testing of tight-fitting respiratory protective equipment in the prior art, complex testing process for users and occupational health-related departments, and inability to be widely, timely and efficiently applied in actual production.
[0004] In a first aspect, a non-invasive, tight-fitting respiratory protective equipment fit test method is provided, comprising:
[0005] Constructing a training sample library for fit testing: collecting head and facial data of the workers and data on the tight-fitting respiratory protective equipment they wear; determining a test action for fit data collection based on the head and facial data and the tight-fitting respiratory protective equipment, and measuring fit level data under the test action, thereby constructing a training sample library for fit testing;
[0006] Suitability level prediction training process: uniformly selecting training data from the training sample library, normalizing the selected training data, and clustering and dividing the data into several groups of training data; training the suitability level prediction model for each group of training data;
[0007] Suitability level prediction test process: Based on a given unknown worker, tight-fitting respiratory protective equipment with different parameters, and test actions, a test sample library is constructed. The data in the test sample library is sent to the trained suitability level prediction model to determine whether the worker's facial shape information matches the respiratory equipment.
[0008] In the above solution, optionally, the constructing of a fitness test training sample library specifically includes:
[0009] Collect the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear, where the head and face data include: face length, face width, tragus-nasal root length, forehead arc length between tragus points, subnasal arc length between tragus points, jaw arc length between tragus points, nose width, and nose height; the tight-fitting respiratory protective equipment data include: length, width, nose clip length / arc length, ear strap / head strap length; the test actions for determining the suitability of data collection include: normal breathing, deep breathing, turning the head left and right, raising and lowering the head, speaking, and bending over;
[0010] According to the standards of quantitative fitness test, fitness level data is measured under different test actions, so as to construct a training sample library for fitness test.
[0011] In the above solution, optionally, the training suitability level prediction model includes:
[0012] Training data is selected from the fit test training sample library to ensure that the training data covers head and facial data of different types of workers and data on tight-fitting respiratory protective equipment. The data selection principles are: the test actions are diverse and the number is evenly distributed; the training data size falls within the preset range to ensure the effectiveness of model training;
[0013] The selected training data are normalized and clustered into several groups of training data; for different groups, multiple suitability level prediction model trainings are carried out respectively.
[0014] In the above solution, further optionally, the training data is normalized using L2 norm, and the normalized data is divided into several groups using K-means clustering method.
[0015] In the above solution, optionally, the suitability level prediction model is a support vector regression model, a random forest model, a K-nearest neighbor algorithm or an artificial neural network model.
[0016] In the above solution, optionally, the test suitability level prediction model includes:
[0017] A test sample library is constructed by providing facial data of an unknown worker, data of a close-fitting respiratory protective equipment with different parameters, and test actions. The test sample library is fed into multiple fitness level prediction training models to obtain prediction results under multiple different fitness test models, and the prediction results are sorted.
[0018] Perform weighted calculation based on the sorting results and output the prediction results, that is, calculate the RMSE weight of each model and output the final prediction results.
[0019] In a second aspect, a non-invasive, tight-fitting respiratory protective equipment fit testing system is provided, the system comprising:
[0020] A module for constructing a training sample library for fit testing is used to collect head and facial data of the workers and the data of the tight-fitting respiratory protective equipment they wear; determine a test action for collecting fit data based on the head and facial data and the tight-fitting respiratory protective equipment, and measure fit level data under the test action, thereby constructing a training sample library for fit testing;
[0021] The fitness level prediction model training process module is used to uniformly select training data from the training sample library, normalize the selected training data, and cluster and divide them into several groups of training data; and perform fitness level prediction model training on each group of training data;
[0022] The fitness level prediction model test process module is used to build a test sample library based on given unknown workers, tight-fitting respiratory protective equipment with different parameters, and test actions. The data in the test sample library is sent to the trained fitness level prediction model to determine whether the worker's facial information matches the respiratory equipment.
[0023] In a third aspect, a computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.
[0024] Compared with the prior art, this application has at least the following beneficial effects:
[0025] Based on further analysis and research of existing technical problems, this application recognizes that the cost of existing respiratory protective equipment fit testing is high, and the testing process for users, production units, or occupational health departments is complex, making it difficult to widely, timely, and efficiently apply it to actual production. By constructing a training sample library, training a fit level prediction model, and conducting inspection tests, this application provides an efficient, low-cost, non-invasive fit testing method for tight-fitting respiratory protective equipment. This method can significantly reduce inspection costs, improve inspection efficiency, and be widely used in actual production to effectively ensure the respiratory safety of workers.
[0026] This application uses a non-invasive method that does not require complex equipment and professionals, significantly reducing the cost of fitness testing. By building a training sample library and training model, it achieves automated fitness testing, reducing manual operation and time costs.
[0027] This application uses multiple machine learning models for prediction and combines the results of multiple models through weighted calculations, improving the accuracy and reliability of the test. The testing process is simple and fast, suitable for widespread application in actual production.
[0028] This application collects diverse head and facial data and tight-fitting respiratory protective equipment parameters to ensure the model is adaptable to different types of workers and equipment. The diverse testing actions enable a comprehensive assessment of the suitability of tight-fitting respiratory protective equipment in different work scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 A schematic flow chart of a non-invasive, tight-fitting respiratory protective equipment fit testing method provided in the first embodiment of the present application.
[0030] Figure 2 A flowchart of a non-invasive, tight-fitting respiratory protective equipment suitability test method provided in the second embodiment of the present application.
[0031] Figure 3 A block diagram of the module architecture of a non-invasive, tight-fitting respiratory protective equipment fit testing system provided in one embodiment of the present application. DETAILED DESCRIPTION
[0032] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.
[0033] In the description of this application: unless otherwise specified, "a plurality of" means two or more. Expressions such as "include", "comprising", "having" and the like also mean "not limited to" (certain units, components, materials, steps, etc.).
[0034] With the widespread application of artificial intelligence technology, artificial intelligence technology represented by machine learning has the characteristics of accuracy, efficiency, strong practicality and wide application range. In terms of fit testing, when applying the paradigm of machine learning for prediction, it can overcome the difficulties of equipment installation, high testing costs, and large number of tests in some complex scenarios, and quickly screen and lock the selection range of tight-fitting respiratory protective equipment. This application intends to introduce a model classification method in the field of machine learning, aimed at the task of predicting the level of fit of tight-fitting respiratory protective equipment, to assist wearers in making choices, improve the level of protection, and prevent the occurrence of respiratory occupational diseases in workers. This will solve the practical problems of high cost, low efficiency, and difficulty in application of existing tight-fitting respiratory protective equipment fit testing.
[0035] In one embodiment, reference Figure 1 and Figure 2 , provides a non-invasive fit test method for tight-fitting respiratory protective equipment, including:
[0036] Construct a training sample library for fit testing: Collect the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear to determine the fit measurement range; determine the test action for fit data collection based on the head and face data and the tight-fitting respiratory protective equipment, and measure the fit level data under the test action to construct a training sample library for fit testing;
[0037] The training process of suitability level prediction: uniformly select training data from the training sample library, normalize the selected training data, and cluster them into several groups of training data; train the suitability level prediction model for each group of training data;
[0038] The test process for suitability level prediction: Based on a given unknown worker, tight-fitting respiratory protective equipment with different parameters, and test actions, a test sample library is constructed. The data in the test sample library is sent to the trained suitability level prediction model to determine whether the worker's facial shape information matches the respiratory equipment.
[0039] In order to solve the problems of difficulty in setting up, low efficiency and high cost of the fit test equipment for the close-fitting respiratory protective equipment of relevant workers, the present application provides a non-invasive fit test method for the close-fitting respiratory protective equipment. Specifically: first, the head and face data of the workers are collected, and combined with the data of the close-fitting respiratory protective equipment they wear, the fit level data is determined by determining the test action for the fit data collection, and a training sample library for the fit test is constructed; secondly, a weighted fit level prediction model is designed, and model training is performed on the basis of the above-constructed fit test training sample library; finally, a test sample library is constructed for a given unknown worker and close-fitting respiratory protective equipment with different parameters and test actions, and the data in the test sample library are used to identify the fit level test results according to the trained fit level prediction model to determine whether the worker's facial information matches the respiratory equipment.
[0040] The non-invasive, tight-fitting respiratory protective equipment fit test method of the present application innovates the existing route of tight-fitting respiratory protective equipment fit test, and will significantly improve the prominent problems of difficult equipment setup, low efficiency, and high cost in the existing fit test process for operators. By fully exploring machine learning methods in the field of artificial intelligence and adapting them to the field of fit test for tight-fitting respiratory protective equipment, the fit test of operators' tight-fitting respiratory protective equipment is realized, which significantly improves the efficiency of fit test for operators' tight-fitting respiratory protective equipment and enhances the user experience of fit test.
[0041] In one embodiment, constructing a fitness test training sample library specifically includes:
[0042] Collect data on the head and face of the workers and the tight-fitting respiratory protective equipment they wear. Head and face data include: face length, face width, tragus-to-nasal root length, forehead arc length between tragus points, subnasal arc length between tragus points, jaw arc length between tragus points, nose width, and nose height. Tight-fitting respiratory protective equipment data include: length, width, nose clip length / arc length, and ear strap / head strap length. Actions to determine suitability for data collection include: normal breathing, deep breathing, turning the head left and right, raising and lowering the head, speaking, and bending over.
[0043] According to the standards of quantitative fitness test, fitness level data is measured under different test actions, so as to construct a training sample library for fitness test.
[0044] In this embodiment, the construction of the fitness test training sample library consists of three parts. The first is to collect the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear, wherein: the head and face data include: morphological face length, face width, tragus and nasal root length, forehead arc length between tragus points, subnasal arc length between tragus points, jaw arc length between tragus points, nose width, and nose height; the tight-fitting respiratory protective equipment data include: length, width, nose clip length / arc length, and ear (head) strap length. The second is to determine the test actions for fitness data collection, including: normal breathing, deep breathing, turning the head left and right, raising and lowering the head, speaking, and bending over. The third is to measure the fitness level data under different test actions according to the standards of quantitative fitness test, thereby constructing a training sample library for fitness test.
[0045] In this embodiment, the fitness test training sample library is composed of three parts: collecting the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear, determining the inspection action for collecting the fitness data, and measuring the fitness level data. Collecting the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear is the prerequisite for constructing the fitness test training sample library; determining the inspection action for collecting the fitness data is to set specific different inspection actions based on the type of tight-fitting respiratory protective equipment, and under this inspection action, obtain the fitness level data based on the measured fitness level data, and finally form the training sample library.
[0046] This example collects comprehensive head and face data and close-fitting respiratory protective equipment data, establishes a variety of testing actions, and performs data collection and measurement in accordance with relevant standards. The resulting fit-testing training sample library is comprehensive, dynamic, standardized, and reliable. This not only improves the model's performance and generalization capabilities, but also significantly enhances the practical application value of fit-testing, effectively ensuring workers' respiratory safety.
[0047] In one embodiment, the training process of the fitness level prediction model includes:
[0048] Training data is selected from the fit test training sample library to ensure that the training data covers head and facial data of different types of workers and data on tight-fitting respiratory protective equipment. The data selection principles are: diversified test actions and evenly distributed numbers; the training data size falls within the preset range to ensure the effectiveness of model training;
[0049] The selected training data are normalized and clustered into several groups of training data; for different groups, multiple suitability level prediction model trainings are carried out respectively.
[0050] In one embodiment, the training data is normalized using L2 norm, and the normalized data is divided into several groups using K-means clustering method.
[0051] In one embodiment, the fitness level prediction model is a support vector regression model, a random forest model, a K-nearest neighbor algorithm, an artificial neural network model, etc.
[0052] In this embodiment, the weighted fitness level prediction model consists of two parts. The first part is to select model training data based on samples in the fitness test training sample library. The principle is to ensure that the types of the above-mentioned tight-fitting respiratory protective equipment are diverse, and the number of samples under different test actions is evenly distributed; the second part is to use the L2 norm to normalize the data required for the above-mentioned modeling, and use the K-means method to cluster it and divide it into several groups; for different groups, multiple fitness level prediction model trainings are carried out respectively.
[0053] The training data selection utilizes data selection in the construction of the training set, following the above-mentioned head and facial data of different types of workers and different types of tight-fitting respiratory protective equipment, so that the inspection actions for suitability data collection are diversified, the number is evenly distributed, and the basic principle that the scale of the selected model training data is above a certain range is met; the suitability level prediction model training is carried out on the basis of the model training data selection, firstly, the L2 norm is used to normalize the data required for the above-mentioned modeling, and the K-means method is used to cluster it and divide it into several groups. For different groups, multiple suitability level prediction model trainings are carried out respectively, where the suitability level prediction model can be a support vector regression model, a random forest model, a K nearest neighbor algorithm, an artificial neural network model, etc.
[0054] Through the above steps, this embodiment can effectively train a fitness level prediction model, ensuring that the model has good generalization ability and prediction performance, thereby providing reliable technical support for the fitness testing of tight-fitting respiratory protective equipment.
[0055] In one embodiment, the testing process of the fitness level prediction model includes:
[0056] The given unknown worker's facial data, the data of the tight-fitting respiratory protective equipment with different parameters, and the test actions are constructed into a test sample library. The test sample library is sent to multiple fitness level prediction training models to obtain the prediction results under multiple different fitness test models, and the prediction results are ranked.
[0057] Perform weighted calculation based on the sorting results and output the prediction results, that is, calculate the RMSE weight of each model and output the final prediction results.
[0058] In this embodiment, the output of the test results of the suitability level is to use the trained model to send the given unknown workers and the tight-fitting respiratory protective equipment with different parameters and the test actions into the aforementioned multiple suitability level prediction training models to obtain the prediction results under multiple different suitability test models, and sort the results. The weighted calculation is performed based on the sorting results, that is, the RMSE weight of each model is calculated; for the sample to be predicted, the prediction results of multiple models are combined with the weight of each model to calculate and output the final prediction result.
[0059] The fitness level test consists of two parts: fitness prediction under multiple different fitness test models and weighted calculation output prediction results. Fitness prediction under multiple different fitness test models is done by using a trained model. The given unknown worker and the tight-fitting respiratory protective equipment with different parameters and the test actions are fed into the aforementioned multiple fitness level prediction training models to obtain the prediction results under the multiple different fitness test models and sort the results. The weighted calculation output prediction results is based on the above sorting results. That is, the RMSE weight of each model is calculated. For the sample to be predicted, the prediction results of multiple models are combined with the weight of each model to calculate and output the final prediction result.
[0060] Through the above steps, this embodiment can effectively perform a test on the suitability level and ensure that the prediction results have good accuracy and stability.
[0061] In one embodiment, reference Figure 3 , provides a non-invasive, tight-fitting respiratory protective equipment fit testing system, including the following program modules: a fit test training sample library construction module, a fit level prediction model training module, and a fit level prediction model testing module, wherein:
[0062] Construct a fit test training sample library module to collect the head and face data of the workers and the tight-fitting respiratory protective equipment they wear; determine the test action for fit data collection based on the head and face data and the tight-fitting respiratory protective equipment, and measure the fit level data under the test action, thereby constructing a training sample library for fit test;
[0063] The fitness level prediction model training process module is used to uniformly select training data from the training sample library, normalize the selected training data, and cluster them into several groups of training data; and conduct fitness level prediction model training on each group of training data;
[0064] The fitness level prediction model test process module is used to build a test sample library based on given unknown workers, tight-fitting respiratory protective equipment with different parameters, and test actions. The data in the test sample library is sent to the trained fitness level prediction model to determine whether the worker's facial information matches the respiratory equipment.
[0065] The specific implementation content of each module can be found in the above definition of a non-invasive, tight-fitting respiratory protective equipment suitability test method, which will not be repeated here.
[0066] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method of the above embodiment.
[0067] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
Claims
1. A non-invasive method for testing the suitability of tight-fitting respiratory protective equipment, characterized in that: include: Constructing a training sample library for fit testing: collecting head and facial data of the workers and the data of the tight-fitting respiratory protective equipment they wear; determining a test action for collecting fit level data based on the head and facial data and the type of tight-fitting respiratory protective equipment, and measuring the fit level data under the test action, thereby constructing a training sample library for fit testing; The training process of suitability level prediction is as follows: uniformly selecting training data from the training sample library, normalizing the selected training data, and clustering and dividing the data into several groups of training data; training the suitability level prediction model for each group of training data; The test process for suitability level prediction: Based on the given unknown workers, the types of tight-fitting respiratory protective equipment with different parameters, and the test actions, a test sample library is constructed. The data in the test sample library is sent to the trained suitability level prediction model to determine whether the worker's facial shape information matches the respiratory equipment.
2. The non-invasive, tight-fitting respiratory protective equipment suitability test method according to claim 1, characterized in that: The construction of the fitness test training sample library specifically includes: Collect the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear, where the head and face data include: face length, face width, tragus-nasal root length, forehead arc length between tragus points, subnasal arc length between tragus points, jaw arc length between tragus points, nose width, and nose height; the tight-fitting respiratory protective equipment data include: length, width, nose clip length / arc length, ear strap / head strap length; the test actions for determining the suitability of data collection include: normal breathing, deep breathing, turning the head left and right, raising and lowering the head, speaking, and bending over; According to the standards of quantitative fitness test, fitness level data is measured under different test actions, so as to construct a training sample library for fitness test.
3. The non-invasive, tight-fitting respiratory protective equipment suitability test method according to claim 1, characterized in that: The training process of the fitness level prediction model includes: Training data is selected from the fit test training sample library to ensure that the training data covers head and facial data of different types of workers and data on tight-fitting respiratory protective equipment. The data selection principles are: the test actions are diverse and the number is evenly distributed; the training data size falls within the preset range to ensure the effectiveness of model training; The selected training data are normalized and clustered into several groups of training data; for different groups, multiple suitability level prediction model trainings are carried out respectively.
4. The non-invasive, tight-fitting respiratory protective equipment suitability test method according to claim 3, characterized in that: The training data is normalized using L2 norm, and the normalized data is divided into several groups using K-means clustering method.
5. The non-invasive, tight-fitting respiratory protective equipment suitability test method according to claim 3, characterized in that: The fitness level prediction model is a support vector regression model, a random forest model, a K-nearest neighbor algorithm or an artificial neural network model.
6. The non-invasive, tight-fitting respiratory protective equipment suitability test method according to claim 1, characterized in that: The testing process of the fitness level prediction model includes: A test sample library is constructed by providing facial data of an unknown worker, data of a close-fitting respiratory protective equipment with different parameters, and test actions. The test sample library is fed into multiple fitness level prediction training models to obtain prediction results under multiple different fitness test models, and the prediction results are sorted. Perform weighted calculation based on the sorting results and output the prediction results, that is, calculate the RMSE weight of each model and output the final prediction results.
7. A non-invasive, tight-fitting respiratory protective equipment fit testing system, characterized in that: include: Construct a fitness test training sample library module to collect the head and face data of the workers and the data of the tight-fitting respiratory protective equipment they wear; Determining a test action for collecting fit data based on the head and face data and the type of tight-fitting respiratory protective equipment, and measuring fit level data under the test action, thereby constructing a training sample library for fit testing; The training process module of the suitability level prediction model is used to uniformly select training data from the training sample library, normalize the selected training data, and cluster and divide them into several groups of training data; and perform suitability level prediction model training on each group of training data; The test process module of the fitness level prediction model is used to build a test sample library based on given unknown workers, tight-fitting respiratory protective equipment with different parameters, and test actions. The data in the test sample library is sent to the trained fitness level prediction model to determine whether the worker's facial information matches the respiratory equipment.
8. A computer device comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method according to claim 1.