Cooling prediction system of pressure steam sterilization bag
Through multi-sensor monitoring and machine learning models, the cooling time of pressure steam sterilization packages is automatically predicted, which solves the problem of low efficiency of manual detection in existing technologies and realizes real-time, accurate prediction and cost optimization of the cooling time of sterilization packages.
Patent Information
- Application Number
- CN202510706682.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-28
- Publication Date
- 2025-09-19
AI Technical Summary
In the existing technology, the prediction of the cooling time of pressure steam sterilization packages relies on manual inspection, which is inefficient and time-consuming, and cannot meet the distribution requirements of the disinfection supply center.
Multi-sensor monitoring of the environmental parameters surrounding the sterilization package is used, and combined with data preprocessing, feature engineering and XGBoost model, automatic prediction of the sterilization package cooling time is achieved, and the prediction accuracy is improved through the model update mechanism.
It realizes the automatic and real-time prediction of the cooling time of the sterilization package, reduces the manual waiting time, improves the accuracy of the prediction and the stability of the system, and reduces the operating costs.
Smart Images
Figure SMS_1 
Figure SMS_2 
Figure SMS_3
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of medical equipment detection, and in particular relates to a temperature drop prediction system for a pressure steam sterilization package. Background Art
[0002] Pressure steam sterilization is the preferred sterilization method in hospitals. The wet pack after sterilization is a key link in the quality control of the disinfection supply center. In order to avoid wet packs, the relevant industry standards of the National Disinfection Supply Center stipulate that the cooling time of items unloaded from the sterilizer is often more than 30 minutes, that is, they can only be distributed when they are close to room temperature.
[0003] In actual work, staff often use hand touch or infrared thermometers to test temperature. Because they are not sure how long it takes for different pressure steam sterilization packages to reach room temperature or close to room temperature, staff often measure it at intervals. The work is tedious and time-consuming, and the disposable rate is extremely low. When the workload is heavy, most staff abandon it, and the distribution requirements of pressure steam sterilization packages cannot be met.
[0004] Therefore, a system is needed that can automatically predict the cooling time of pressure steam sterilization packages after sterilization treatment. By monitoring parameters of different sensors, it can automatically predict the cooling time of pressure steam sterilization packages, making it easier for medical staff to pick them up and use them, reducing waiting time and manual inspection costs. Summary of the Invention
[0005] In view of the above shortcomings, the present invention provides a temperature drop prediction system for pressure steam sterilization packages, which includes the following modules:
[0006] A data acquisition module for collecting environmental parameters of pressure steam sterilization packages;
[0007] A data preprocessing module that filters and cleans the collected parameters;
[0008] Feature engineering module for extracting features from preprocessed parameters;
[0009] A prediction model module for predicting the cooling time of pressure steam sterilization packages based on the extracted features;
[0010] An update module that dynamically adjusts model parameters in the prediction model module by monitoring the difference between the predicted and actual cooling conditions;
[0011] and communication and storage modules for data storage and transmission.
[0012] Furthermore, the data acquisition module monitors the environmental parameters around the pressure steam sterilization package through a temperature sensor, a humidity sensor and an air flow rate sensor;
[0013] Based on the monitored data, the temperature change rate is calculated using the following formula:
[0014]
[0015] Among them, T current is the temperature of the current minute, T previous is the temperature in the previous minute, and Δt is the time interval.
[0016] Furthermore, every three of the temperature sensors, humidity sensors and air flow rate sensors are arranged as a group and are set with independent identity ID labels.
[0017] Furthermore, the data preprocessing module sequentially performs data filtering, outlier detection and data formatting, and the data filtering adopts a moving average filtering method;
[0018] Outlier detection uses the Z-score method.
[0019] Furthermore, the features extracted by the feature engineering module include basic features, material type features, time series features, and interaction features;
[0020] The interaction features are standardized and normalized to scale the features to the range of [0, 1].
[0021] Furthermore, the XGBoost model is used as the main regression model in the prediction model module, and its training process is as follows:
[0022] 1) Use the preprocessed and feature-engineered data as a dataset;
[0023] 2) Divide the dataset into training set, validation set and test set;
[0024] 3) Model training through the XGBoost library;
[0025] 4) Optimize the hyperparameters of the trained XGBoost model;
[0026] 5) Use mean squared error and mean absolute error to evaluate the performance of the XGBoost model;
[0027] 6) Save the trained XGBoost model for future use.
[0028] Compared with the prior art, the present invention has the following beneficial effects:
[0029] 1. By integrating multiple key parameters such as ambient temperature, surface temperature, humidity, and air velocity, it ensures data comprehensiveness while reducing the manual waiting time and operating costs of medical staff;
[0030] 2. It has real-time data processing capabilities. Through the data pre-processing module, data filtering, outlier detection, and instant calculation of temperature change rates, it ensures the quality and timeliness of data input to the prediction model module. This real-time performance not only improves the prediction response speed, but also dynamically reflects environmental changes, helping users to adjust operating parameters in a timely manner and achieve more refined management and control.
[0031] 3. The feedback and model update mechanism can automatically trigger model optimization and update by monitoring the error between the actual cooling time and the predicted results, ensuring that the prediction model module always maintains high accuracy, significantly improving the long-term stability and adaptability of the system.
[0032] In summary, by introducing advanced machine learning models, real-time data processing, optimizing user interaction experience, and enhancing the system's adaptability and remote management functions, the system significantly outperforms the existing pressure steam sterilization package cooling time prediction system in terms of accuracy, cost-effectiveness, operational convenience, and system flexibility, demonstrating higher practical value and market competitiveness. DETAILED DESCRIPTION
[0033] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.
[0034] Example
[0035] This embodiment provides a temperature drop prediction system for pressure steam sterilization packages, including a data acquisition module, a data preprocessing module, a feature engineering module, a prediction model module, an update module, and a communication and storage module, as follows:
[0036] The data acquisition module monitors the environmental parameters around the pressure steam sterilization package through temperature sensors, humidity sensors, and air flow rate sensors. The temperature sensors include a digital temperature sensor (using DS18B20) to monitor the temperature of the pressure steam sterilization package cooling environment, and a non-contact infrared temperature sensor (using MLX90614) to monitor surface temperature changes in the area around the pressure steam sterilization package. The humidity sensor is used to measure the humidity of the pressure steam sterilization package cooling environment (using SHT31), and the air flow rate sensor is used to measure the air flow rate of the pressure steam sterilization package cooling environment (using Honeywell AWM series).
[0037] Based on the monitored data, the temperature change rate can also be calculated:
[0038] The temperature change rate is calculated as:
[0039]
[0040] Among them, T current is the temperature of the current minute, T previous is the temperature of the previous minute, Δt is the time interval (1 minute);
[0041] The data collected from the pressure steam sterilization package of a stainless steel instrument within a certain period of 10 minutes is as follows:
[0042] As shown in Table 1:
[0043]
[0044] Table 1
[0045] Substituting the temperature at the corresponding time into the above formula can obtain the corresponding temperature change rate:
[0046] For example, in the 6th minute:
[0047] It should be noted that the above three sensors form a group (multiple groups are provided) and use a unified ID to facilitate subsequent statistics; each sensor is installed at the corresponding monitoring position of the pressure steam sterilization package rack. The pressure steam sterilization package rack is a multi-layer rack structure, and a group of sensors is placed on each layer. Since each group of sensors corresponds to a pressure steam sterilization package and has an independent ID, the data will not interfere with each other during subsequent processing. Only one digital temperature sensor can also be used to monitor the temperature in the cooling chamber.
[0048] In addition, since the contents of the pressure steam sterilization package include dressings, stainless steel instruments, plastics, etc., as well as packaging materials, each pressure steam sterilization package rack needs to be classified according to each layer to facilitate the subsequent corresponding different processing processes.
[0049] The data preprocessing module processes the data collected by the data acquisition module (processing each group of IDs separately):
[0050] First, data filtering is performed. In this embodiment, a moving average filtering method is used to apply moving average filtering to the data of sensors such as ambient temperature (T_env), surface temperature (T_surface), humidity (H), and air velocity (V). The window size is 3 minutes. For example, the ambient temperature is calculated as follows:
[0051] Import the ambient temperatures at minutes 4, 5, and 6, 24.8°C, 24.9°C, and 25.0°C, respectively, into the formula:
[0052]
[0053] Among them, y n is the value after filtering, x n -i is the raw data of minute ni, i.e., the current data point and its previous N-1 data points. N is the window size, which indicates the number of consecutive data points used to calculate the average value (e.g., 4, 5, and 6 minutes represent 3 windows, i.e., N=3 at this time).
[0054] get:
[0055]
[0056] Similarly, other sensor data are also filtered in the same way, which will not be described here.
[0057] Then, outlier detection is performed to identify and remove outliers to ensure data quality and improve the robustness of the model. This embodiment uses the Z-score method to calculate the Z-score of each data point. If the absolute value exceeds 3, it is considered an outlier. The specific formula is as follows:
[0058]
[0059] Where x is the current data point, μ is the mean within the window, and σ is the standard deviation within the window;
[0060] The ambient temperature (T_env) at the 6th minute is 25.0°C, and the data in the window are 24.7°C, 24.8°C, and 24.9°C. After importing, we get:
[0061]
[0062] Then you can get the Z-score:
[0063]
[0064] Since Z=2.44<3, it is proved that the data point is a valid value.
[0065] Finally, data formatting is performed to unify the cleaned, filtered, and calculated data into a unified format for subsequent processing, as shown in Table 2 below:
[0066]
[0067] Table 2
[0068] The feature engineering module extracts predictive features from preprocessed data to improve the model's expressiveness. The features extracted include:
[0069] 1) Basic characteristics, namely ambient temperature (T _ env), surface temperature (T _surface), humidity (H), air velocity (V), and cooling start time (t);
[0070] 2) Material type characteristics (such as stainless steel, plastic, etc.), and encode the material category through One-Hot Encoding;
[0071] 3) Time series characteristics, i.e., temperature change rate;
[0072] 4) Interaction features, i.e. constructing interaction terms between features, such as the product of the average ambient temperature and humidity;
[0073] Also use the data from the 6th minute: ambient temperature (T _ env) is 24.9℃, surface temperature (T _ The surface temperature is 24.0°C, the humidity (H) is 60%, the air velocity (V) is 5.0 m / s, and the cooling start time (t) is 10 minutes. It can be concluded that the temperature change rate is 0.1°C / minute, and the interaction characteristics are:
[0074] Interaction feature = T env ×H=24.9×60=1494;
[0075] Then, standardization and normalization (using Min-Max normalization) are performed to scale the features to the range of [0, 1] to improve the training efficiency and performance of the model:
[0076] The scope of each feature is formulated as follows:
[0077] T_env is 20℃30℃, T_surface is ℃30℃, H is 40%80%, V is 0m / s-10m / s, ΔT_env is -1℃ / min-1℃ / min, and the interaction characteristics are 800-2000;
[0078]
[0079] Where X is the original data value, X min is the minimum value of the data, X max is the maximum value of the data;
[0080]
[0081] Normalized eigenvector = {0.49, 0.48, 0.6, 0.5, 0.05, 0.05, 0, 1, 0, 0, 0.747};
[0082] That is, we get the complete standardized and normalized feature vector:
[0083] feature Normalized value Ambient temperature (T_env) 0.49 Surface temperature (T_surface) 0.48 Humidity (H) 0.6 Air velocity (V) 0.5 Feature Interaction 0.747
[0084] Table 3
[0085] The prediction model module uses a machine learning model to predict the cooling time of the pressure steam sterilization package based on the extracted features.
[0086] The machine learning model in this embodiment uses the XGBoost model as the main regression model, which performs well in processing nonlinear relationships and feature importance evaluation. The training process is as follows:
[0087] 1) Use the preprocessed and feature-engineered data in Table 2 as the dataset;
[0088] 2) Divide the dataset into training set (70%), validation set (15%) and test set (15%);
[0089] 3) Model training through the XGBoost library;
[0090] 4) Optimize the hyperparameters of the XGBoost model to improve model performance (you can use GridSearchCV or RandomizedSearchCV for hyperparameter optimization);
[0091] 5) Use mean squared error (MSE) and mean absolute error (MAE) to evaluate model performance and ensure that the errors on the validation set and test set are within an acceptable range;
[0092] 6) Save the trained XGBoost model for future use.
[0093] Still using the data from the 6th minute: ambient temperature (T_env) is 24.9℃, surface temperature (T_surface) is 24.0℃, humidity (H) is 60%, air velocity (V) is 5.0m / s and cooling start time (t) is 10 minutes, it can be concluded that the temperature change rate is 0.1℃ / minute as data, which is substituted into the XGBoost model;
[0094] Based on the feature vector {0.49, 0.48, 0.6, 0.5, 0.05, 0.05, 0, 1, 0, 0, 0.747}, the predicted cooling time is 39.5 minutes.
[0095] The update module can dynamically adjust and optimize the model based on the error between the actual cooling time and the predicted result to improve the prediction accuracy. The specific steps include:
[0096] 1) Obtain the actual cooling time through manual recording or other equipment;
[0097] 2) Calculation error:
[0098]
[0099] Where e is the prediction error, which represents the absolute error between the predicted cooling time and the actual cooling time. To predict the cooling time, that is, the time required for the pressure steam sterilization package to cool down as predicted by the XGBoost model, The actual cooling time is the time actually measured or recorded for the pressure steam sterilization package to complete cooling.
[0100] It should be noted that if e>δ (preset threshold, such as 5 minutes, which can be set according to actual conditions), the model update rule is triggered, for example:
[0101] e=|39.5-41.0|=1.5 minutes, then e=1.5<5, which means that the current prediction error is within an acceptable range and there is no need to update the model immediately.
[0102] 3) Based on the prediction error, adjust the model’s hyperparameters or select a more appropriate feature combination;
[0103] 4) Record each error and adjustment details to facilitate subsequent analysis and optimization.
[0104] The communication and storage module is used for wireless transmission and storage of data, which is convenient for remote monitoring and centralized management (the ESP32 processor is used in this embodiment, with a built-in WIFI module for wireless data transmission and basic data processing). At the same time, a cloud server can be set up as needed to store historical data and model parameters (a local database can be used conventionally). It should be noted that, considering the computing requirements of the complex XGBoost model, it is also necessary to combine a more powerful processor (such as Raspberry Pi) to achieve high-precision real-time prediction, which can be integrated into the background server terminal.
[0105] It should also be noted that the predicted time information can be output to the display screen inside the department through the background server terminal, so that medical staff can observe the time when cooling is completed and the product can be taken away.
[0106] It should be noted that the structure described in the present invention can be implemented in a variety of different forms and is not limited to the described embodiments. Any equivalent transformations made by ordinary technicians in this field using the contents of the present invention specification, or directly or indirectly applied to other related technical fields, such as the loading and unloading of other items, are included in the scope of protection of the present invention.
Claims
1. A temperature drop prediction system for pressure steam sterilization packages, characterized in that: Includes the following modules: A data acquisition module for collecting environmental parameters of pressure steam sterilization packages; A data preprocessing module that filters and cleans the collected parameters; Feature engineering module for extracting features from preprocessed parameters; A prediction model module for predicting the cooling time of pressure steam sterilization packages based on the extracted features; An update module that dynamically adjusts model parameters in the prediction model module by monitoring the difference between the predicted and actual cooling conditions; and communication and storage modules for data storage and transmission.
2. The temperature drop prediction system for a pressure steam sterilization package according to claim 1, wherein: The data acquisition module monitors the environmental parameters around the pressure steam sterilization package through a temperature sensor, a humidity sensor and an air flow rate sensor; Based on the monitored data, the temperature change rate is calculated using the following formula: Among them, T current is the temperature of the current minute, T previous is the temperature in the previous minute, and Δt is the time interval.
3. The temperature drop prediction system for a pressure steam sterilization package according to claim 2, wherein: Every three of the temperature sensors, humidity sensors and air flow rate sensors are arranged into a group and are set with independent identity ID labels.
4. The temperature drop prediction system for a pressure steam sterilization package according to claim 1, wherein: The data preprocessing module sequentially performs data filtering, outlier detection and data formatting, and the data filtering adopts a moving average filtering method; Outlier detection uses the Z-score method.
5. The temperature drop prediction system for a pressure steam sterilization package according to claim 1, wherein: The features extracted by the feature engineering module include basic features, material type features, time series features and interaction features; The interaction features are standardized and normalized so that the interaction features are scaled to the range of [0, 1].
6. The temperature drop prediction system for a pressure steam sterilization package according to claim 1, wherein: The XGBoost model is used as the main regression model in the prediction model module, and its training process is as follows: 1) Use the preprocessed and feature-engineered data as a dataset; 2) Divide the dataset into training set, validation set and test set; 3) Model training through the XGBoost library; 4) Optimize the hyperparameters of the trained XGBoost model; 5) Use mean squared error and mean absolute error to evaluate the performance of the XGBoost model; 6) Save the trained XGBoost model for future use.