Time point method for critical point of sterilization period of pressure steam sterilizer

Abstract

The application provides a time point selection method for a critical point of a sterilization period of a pressure steam sterilizer, and comprises the following steps: signal acquisition, intercepting an optimal data processing area for judging the sterilization period in an original signal to obtain a to-be-processed signal, performing smoothing processing on the to-be-processed signal to obtain temperature data after smoothing processing, determining training data and target data of a point selection method, signal fitting, and point selection. The time point selection method for the critical point of the sterilization period of the pressure steam sterilizer has the multiple advantages and beneficial effects of high automation degree, high precision, wide applicability, improved work efficiency, and reduced cost.

Description

Technical Field

[0001] This invention relates to a method for determining the critical point of sterilization in a pressure steam sterilizer. Background Technology

[0002] Monitoring physical parameters during the sterilization process of a pressure steam sterilizer is essential to ensuring sterilization effectiveness and is crucial for medical institutions to prevent cross-infection and major medical accidents. Meanwhile, the national health industry standard WS310.3-2016 stipulates that the effective sterilization period of a pressure steam sterilizer should have a temperature fluctuation range within +3℃, and the sterilization time should meet the minimum sterilization time requirements. Furthermore, the time, temperature, and pressure values ​​at all critical points should be recorded, and the results should meet the sterilization requirements. National standard GB / T 30690—2014 clearly stipulates that the measured sterilization time should not be lower than the set value and should not exceed 10% of the set value. Based on the above, monitoring the physical parameters at the start and end points of the sterilization period is particularly important.

[0003] Existing technologies are mainly based on traditional manual point selection methods, which require manual observation of monitoring signals and selection based on experience and subjective judgment. This approach has the following problems and drawbacks: 1. Low efficiency: Manual observation and judgment of each point is time-consuming and labor-intensive; 2. Low accuracy: Subjective judgment introduces errors and uncertainties, making accurate point selection difficult, with errors reaching up to 40%; 3. High experience requirement: Extensive experience and professional knowledge are required, which is challenging for novice supervisors; 4. Unstable results: Human factors and uncertainties can lead to significant differences and fluctuations in selection results; 5. Difficulty in quantifying point selection: Manual point selection methods are often difficult to quantify and standardize, making statistical analysis and comparison challenging. Summary of the Invention

[0004] This invention provides a method for determining the critical point of sterilization in a pressure steam sterilizer to solve one or more of the above-mentioned problems.

[0005] According to one aspect of the present invention, a method for determining the critical point of sterilization period in a pressure steam sterilizer is provided, comprising the steps of: A. Signal acquisition: acquiring the raw temperature signal of the pressure steam sterilizer through a Bragg fiber grating sensor;

[0006] B. Extract the optimal data processing region from the original signal to determine the sterilization period, and obtain the signal to be processed;

[0007] C. Smooth the signal to be processed to remove noise and obtain smoothed temperature data;

[0008] D. Determine the input data and target data: The input data is the time series of the optimal data processing area for determining the sterilization period, corresponding to the time data of the signal collected by FBG; the target data is the temperature data after processing in step C, corresponding to the time data.

[0009] E. Signal Fitting: Map the input data to the corresponding temperature data processed in step D, and calculate the mean absolute error between the fitted signal and the target signal to quantify the fitting error.

[0010] F. Determine the critical point of the sterilization period: The starting point of the sterilization period is the point where the temperature is first greater than or equal to the temperature threshold of the sterilizer, and the ending point is the point where the temperature is last greater than or equal to the temperature threshold of the sterilizer, and the temperature must be greater than the temperature threshold during the period from the starting point to the ending point.

[0011] The time-determining method for the critical point of sterilization period in a pressure steam sterilizer of the present invention has a high degree of automation, enabling automatic signal smoothing and automatic point selection, greatly reducing the subjectivity and error of manual operation; it has high accuracy, a high fitting coefficient, and small error; it has wide applicability, not only for temperature monitoring but also for other fields requiring point selection, and has broad application prospects; it improves work efficiency and reduces costs, enabling automatic signal processing and automatic point selection, which can greatly shorten the point selection time, improve work efficiency, reduce data processing and point selection time to 1 second, and also reduce the labor cost of manual point selection.

[0012] In summary, the automatic point selection method proposed in this invention has multiple advantages and beneficial effects, such as high degree of automation, high precision, wide applicability, improved work efficiency, and reduced costs.

[0013] In some embodiments, in step B of the present invention, the moment when the original signal first exceeds or equals the sterilizer temperature threshold is set as... The sterilization time set by the sterilizer is set to The optimal data processing area starting point for determining the sterilization period, as set by this invention, is... The optimal data processing region endpoint for determining the sterilization period is... Therefore, the optimal data processing area for determining the sterilization period in this invention is... .

[0014] In some embodiments, the present invention uses the EMD (Empirical Mode Decomposition) algorithm in step C to decompose the signal to be processed obtained in step B, and obtains multiple IMF (Intrinsic Mode Function) components. Then, the IMF components are subjected to wavelet threshold denoising to obtain smoothed temperature data.

[0015] In some embodiments, the present invention employs moving average, weighted average, median filtering, or wavelet transform to smooth the signal to be processed in step C.

[0016] In some embodiments, in step E, the method for signal fitting is to create a feedforward neural network for function fitting; the dataset is divided: the data processed in step C is divided into a training set, a validation set, and a test set, and the neural network is trained.

[0017] In some embodiments, the size of the hidden layer of the neural network of the present invention is 10.

[0018] In some embodiments, 70% of the data processed in step C of the present invention is assigned to the training set, 15% to the validation set, and 15% to the test set.

[0019] In some embodiments, the sensor of the present invention is a Fiber Bragg Grating (FBG) sensor.

[0020] By organically combining various signal processing techniques such as EMD decomposition of the original signal, wavelet threshold denoising, and neural network fitting, the algorithm proposed in this application can achieve high-precision point selection, accurately determine the start and end points of the sterilization period, and has a high fitting coefficient and small fitting error. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a method for determining the critical point of sterilization in a pressure steam sterilizer according to one embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the temperature monitoring device for a time-determining method of the critical point of sterilization period in a pressure steam sterilizer, according to one embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of the original temperature signal of a pressure steam sterilizer obtained according to one embodiment of the present invention.

[0024] Figure 4 From Figure 3 The optimal data processing area for determining the sterilization period is extracted from the data.

[0025] Figure 5(a) shows the... Figure 4 The signal in the image is decomposed using the EMD algorithm to obtain the IMF component image; Figure 5(b) shows the IMF component image after wavelet threshold denoising.

[0026] Figure 6The temperature data is smoothed by superimposing the denoised IMF components in Figure 5(b).

[0027] Figure 7(a) is a schematic diagram of the neural network training process; Figure 7(b) is a schematic diagram of the neural network training state; Figure 7(c) is a schematic diagram of the error bar; Figure 7(d) is a schematic diagram of the degree of fitting; Figure 7(e) is a schematic diagram of the neural network fitting signal.

[0028] Figure 8 A schematic diagram for determining the critical point of sterilization period by fitting the signal graph in Figure 7(e).

[0029] Figure 9 This is a schematic diagram of the original temperature signal obtained from a pressure steam sterilizer according to another embodiment of the present invention.

[0030] Figure 10 From Figure 9 The optimal data processing area for determining the sterilization period is extracted from the data.

[0031] Figure 11(a) shows the... Figure 9 The signal in the image is decomposed using the EMD algorithm to obtain the IMF component map;

[0032] Figure 11(b) shows the IMF components after wavelet threshold denoising.

[0033] Figure 12 The temperature data is smoothed by superimposing the denoised IMF components in Figure 11(b).

[0034] Figure 13(a) is a schematic diagram of the neural network training process; Figure 13(b) is a schematic diagram of the neural network training state; Figure 13(c) is a schematic diagram of the error bar; Figure 13(d) is a schematic diagram of the degree of fitting; Figure 13(e) is a schematic diagram of the neural network fitting signal.

[0035] Figure 14 A schematic diagram for determining the critical point of sterilization period by fitting the signal graph in Figure 13(e).

[0036] Figure 15 This is a schematic diagram of the original temperature signal obtained from a pressure steam sterilizer according to another embodiment of the present invention.

[0037] Figure 16 From Figure 15 The optimal data processing area for determining the sterilization period is extracted from the data.

[0038] Figure 17(a) shows the... Figure 15 The signal in the image is decomposed using the EMD algorithm to obtain the IMF component image; Figure 17(b) shows the IMF component image after wavelet threshold denoising.

[0039] Figure 18 The temperature data is smoothed by superimposing the denoised IMF components in Figure 17(b).

[0040] Figure 19(a) is a schematic diagram of the neural network training process; Figure 19(b) is a schematic diagram of the neural network training state; Figure 19(c) is a schematic diagram of the error bar; Figure 19(d) is a schematic diagram of the fitting degree; Figure 19(e) is a schematic diagram of the neural network fitting signal.

[0041] Figure 20 A schematic diagram for determining the critical point of sterilization period by fitting the signal graph in Figure 19(e).

[0042] Figure 21 The image shows the smoothing effects of various smoothing methods.

[0043] Figure 22 This is a reference chart comparing the smoothing effects of multiple smoothing methods.

[0044] Figure 23 This is a reference graph for evaluating the effects of multiple smoothing methods. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0047] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising" or "including" include not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0048] The present invention will now be described in further detail with reference to the accompanying drawings.

[0049] Figure 1The flowchart of a time-determining method for the critical point of sterilization in a pressure steam sterilizer according to an embodiment of the present invention is illustrated.

[0050] refer to Figure 1 As shown, this fixed-point method includes the following steps:

[0051] A. Signal Acquisition: The raw temperature signal of the pressure steam sterilizer is acquired through sensors. Sensors can be Bragg fiber grating (FBG) sensors or electronic sensors, etc.

[0052] B. Extract the optimal data processing region from the original signal to determine the sterilization period, and obtain the signal to be processed. In this invention, the moment when the original signal first exceeds or equals the sterilizer temperature threshold is set as... The sterilization time set by the sterilizer is set to The optimal data processing area starting point for determining the sterilization period, as set by this invention, is... The optimal data processing region endpoint for determining the sterilization period is... Therefore, the optimal data processing area for determining the sterilization period in this invention is... Within this processing area, the judgment error is small, resulting in minimal systematic error. Therefore, this application improves data processing efficiency by setting data extraction rules to automatically extract the optimal data processing area for judging the sterilization period.

[0053] C. Smooth the signal to be processed to remove noise and obtain smoothed temperature data.

[0054] D. Determine the input data and target data: The input data is the time series of the optimal data processing area for determining the sterilization period, corresponding to the time data of the signals collected by the sensor; the target data is the temperature data after processing in step C, corresponding to the time data.

[0055] E. Signal Fitting: The signal is fitted using a neural network, mapping the input data to the corresponding temperature data processed in step C.

[0056] F. Determine the critical point of sterilization period: The critical point of sterilization period is determined by computer software. The start point of sterilization period is the point when the temperature is first greater than or equal to the temperature threshold of the sterilizer, and the end point is the point when the temperature is last greater than or equal to the temperature threshold of the sterilizer, and the temperature is greater than the temperature threshold during the period from the start point to the end point.

[0057] Figure 2 The structure of a temperature monitoring device for a time-pointing method for determining the critical point of sterilization in a pressure steam sterilizer according to an embodiment of the present invention is illustrated schematically.

[0058] refer to Figure 2As shown, the device includes: a capillary stainless steel armored FBG temperature sensor 11, which is placed in the sterilization chamber of the pressure steam sterilizer 55 to be monitored. The sensor's pigtail 10 is led out through the front closed door of the sterilizer and connected to a fiber optic demodulator 53. The fiber optic demodulator 53 records the temperature value recorded by the sensor in real time during the operation of the pressure steam sterilizer 55 to be monitored.

[0059] Figures 3-8 The diagram schematically illustrates a method for determining the critical point of sterilization in a pressure steam sterilizer according to an embodiment of the present invention.

[0060] Example 1:

[0061] Taking the temperature monitoring data of the medical high-pressure steam sterilizer BKQ-B50II (the sterilizer's temperature threshold, i.e., the sterilizer's rated operating temperature is 134℃, and the sterilization time is set to 6 minutes) as an example, the critical point of its sterilization period is automatically determined.

[0062] The sterilizer has a rated operating temperature of 134°C. For other sterilizers, such as those with a rated operating temperature of 121°C and a sterilization time set to 20 minutes, the time-determining method described in this application is also fully applicable; only the temperature threshold and sterilization time setting need to be adjusted. The time-determining method for the sterilization critical point of this pressure steam sterilizer is applicable to all models of pressure steam sterilizers on the market.

[0063] Step A: Signal Acquisition. A fiber Bragg grating (FBG) sensor is used to acquire the raw temperature signal. In this embodiment, temperature data signals from 3412 sampling points are acquired (sampling once per second). This data contains a large amount of noise, which affects the manual selection of sampling points and can lead to significant errors. The acquired raw signal is referenced below. Figure 3 As shown.

[0064] Step B: Extract the optimal data processing area for determining the sterilization period to obtain the signal to be processed. Extensive repeatable experiments have shown that the systematic error of extending the set sterilization time by 30% before and after selecting the optimal data processing area is small, and the data is intuitive and clear. Subsequently, the processing area selection is defined by setting the program, with the moment in the original signal first exceeding or equaling the sterilizer temperature threshold set as... The sterilization time set by the sterilizer is set to The optimal data processing area starting point for determining the sterilization period, as set by this invention, is... The optimal data processing region endpoint for determining the sterilization period is... Therefore, the optimal data processing area for determining the sterilization period in this invention is... In this embodiment, 30% of the sterilization time is 108 seconds, and the specific sterilization period for the treatment area is 1664-2240 seconds.

[0065] Step C: Smooth the signal to be processed to obtain smoothed temperature data.

[0066] First, the signal in the data processing area is decomposed using EMD. The signal obtained in step B is decomposed using the EMD algorithm to obtain 7 IMF components. The IMF components obtained by the EMD algorithm are shown in Figure 5(a).

[0067] Secondly, noise removal is performed. The IMF components after EMD decomposition are subjected to wavelet thresholding for noise reduction. The basic steps include wavelet decomposition, thresholding, and wavelet reconstruction.

[0068] ① Wavelet Decomposition: This step first decomposes the original signal (IMF component in this example) through wavelet transform into multi-level wavelet detail coefficients and wavelet approximation coefficients at each level. In this embodiment, the wavelet decomposition level is 3, and the wavelet type used is an 8th-order symmetric wavelet. The goal of wavelet decomposition is to analyze the signal at different frequency levels, thereby enabling the identification and removal of noise at different frequency levels.

[0069] ② Thresholding: The purpose of thresholding is to set wavelet coefficients considered as noise (i.e., coefficients less than a certain threshold) to zero. In this embodiment, the threshold function used is the "logarithm squared" threshold function, the threshold strategy is "maximum likelihood estimation," and the threshold type is "soft thresholding." The soft thresholding function can smooth the data and retain more data characteristics, while the maximum likelihood estimation method sets the threshold based on statistical principles to achieve the best denoising effect.

[0070] ③ Wavelet Reconstruction: Wavelet reconstruction is performed using the processed wavelet coefficients to obtain the denoised signal. During the reconstruction process, wavelet coefficients set to zero do not affect the signal, thus removing noise.

[0071] Thresholds are calculated using the following formula. Calculation:

[0072]

[0073] in, It is the standard deviation of the data.

[0074] The soft thresholding rule is executed using the following formula: for each wavelet coefficient, if its absolute value is less than the threshold, it is set to zero; if its absolute value is greater than the threshold, the threshold is subtracted and the sign is retained.

[0075]

[0076] in, It is a symbolic function. These are the wavelet coefficients after thresholding. These are the original wavelet coefficients. This represents the maximum value in the data. It is a threshold.

[0077] In this embodiment, the threshold for

[0078]

[0079] After wavelet threshold denoising of each IMF component, they are superimposed to obtain the signal after noise removal. The smoothed IMF components are shown in Figure 5(b), and the smoothed signal generated by superimposing the smoothed IMF components is shown in Figure 5(b). Figure 6 As shown.

[0080] Step D: Determine the input and target data. The input data of this invention is the time series of the sterilization period data processing area (i.e., x=1664 to x=2240), corresponding to the time data of the signals collected by FBG. The target data is the temperature signal values ​​corresponding to these time points, which are temperature signal data after decomposition and smoothing by applying the EMD algorithm.

[0081] Step E: Fit the signal.

[0082] First, a neural network is created. A feedforward neural network for function fitting is created using computer software. The hidden layer size of this neural network is 10. The size of the hidden layer affects the model complexity of the neural network; too large a size may lead to overfitting, while too small a size may lead to underfitting. Since the sterilization period dataset is relatively complex, this invention determines the number of hidden layers to be 10 based on the signal characteristics of the sterilization period data and experimental results. In other embodiments, different numbers of hidden layers can be set according to the specific data processing volume.

[0083] Secondly, the dataset is divided. In this embodiment, the data processed in step C is divided into a training set, a validation set, and a test set. In this process, 70% of the data is used for training, 15% is used to validate the generalization ability of the neural network to prevent overfitting, and 15% is used for the final test to evaluate the model's performance on unseen data. This ratio was determined based on numerous repetitive experiments; extensive experimental results have demonstrated that this ratio results in fast training speed and good performance, therefore, this ratio was chosen for dividing the dataset.

[0084] Next, the neural network is trained. This embodiment uses computer software to train the neural network, updating the weights and biases based on the backpropagation algorithm. During training, the loss function used is Mean Absolute Error (MAE), which is the average of the absolute values ​​of the errors. It accurately reflects the magnitude of the actual prediction error, has better robustness to outlier data, and reduces the impact of noise on point selection. This is the most commonly used loss function in neural network training. Its calculation formula is:

[0085]

[0086] in, This is expressed as the number of data points (sample size). No. The actual value of the dependent variable for each data point. No. Predicted values ​​of the dependent variable for each data point.

[0087] MAE assesses the degree of deviation between the true value and the predicted value, i.e., the actual magnitude of the prediction error. The smaller the MAE value, the smaller the fitting error and the better the fit.

[0088] Finally, after training, the neural network is able to map the new input (the time series of sterilization data processing area, i.e., x=1664 to x=2240) to the corresponding output (the smoothed signal). This invention quantifies the fitting error by calculating the mean absolute error (MAE) between the fitted signal and the target signal.

[0089] In this embodiment, the goodness of fit is as high as 0.99992, and the fitting error is 0.0528℃. The training data and results are shown in Figure 7. In this embodiment, the network actually iterated 446 times, with an actual gradient of 0.0478, and its damping factor (Mu) has an actual value of 1 * e^(-7).

[0090] Step F: Determine the sterilization period critical point. Set the temperature judgment according to national standards to determine the start and end points. The sterilization period begins when the temperature first exceeds or equals the sterilizer's temperature threshold, and ends when the temperature last exceeds or equals the sterilizer's temperature threshold, provided that the temperature remains above the threshold throughout the period from start to end. In this embodiment, the start time is 1774 seconds (the manually selected end time is 1773 seconds, a difference of 1 second), and the end time is 2150 seconds (the same as the manually selected start time, 2150 seconds). The point determination results are referenced. Figure 8 As shown.

[0091] All the above steps can be executed and the result obtained in just 1 second.

[0092] Figures 9-14 The diagram schematically illustrates a method for determining the critical point of sterilization in a pressure steam sterilizer according to another embodiment of the present invention.

[0093] Example 2:

[0094] Taking the temperature monitoring data of the Sanqiang pre-vacuum rapid pressure steam sterilizer SQ-Y45 (the sterilizer's temperature threshold, i.e., the sterilizer's rated operating temperature is 134℃, and the sterilization time is set to 6 minutes) as an example, the critical point of its sterilization period is automatically determined.

[0095] Step A: Signal Acquisition. A fiber Bragg grating sensor is used to acquire the raw temperature signal. In this embodiment, temperature data signals from 3329 sampling points are acquired (sampling once per second). This data contains a large amount of noise, which affects the manual selection of sampling points and leads to significant errors. (Sensing device reference) Figure 2 As shown, the original signal reference Figure 9 As shown.

[0096] Step B: Extract the optimal data processing area for determining the sterilization period. This invention uses a program to define the selection criteria for the processing area; the moment when the original signal first exceeds or equals the sterilizer temperature threshold is set as... The sterilization time set by the sterilizer is set to The optimal data processing area starting point for determining the sterilization period, as set by this invention, is... The optimal data processing region endpoint for determining the sterilization period is... Therefore, the optimal data processing area for determining the sterilization period in this invention is... In this embodiment, 30% of the sterilization time is 108 seconds. The specific sterilization period data ranges from 1839 to 2415 seconds. The optimal data processing area for determining the sterilization period is selected in this embodiment for reference. Figure 10 As shown.

[0097] Step C: Smooth the signal to be processed to obtain smoothed temperature data.

[0098] First, the signal in the data processing area is decomposed using EMD. The signal obtained in step B is decomposed using the EMD algorithm to obtain 7 IMF components. The IMF components obtained by the EMD algorithm are shown in Figure 11(a).

[0099] Secondly, noise is removed. Wavelet thresholding is performed on each IMF component after EMD decomposition.

[0100] In this embodiment, the threshold for:

[0101]

[0102] After wavelet threshold denoising of each IMF component, they are superimposed to obtain the signal after noise removal. The smoothed IMF components are shown in Figure 11(b), and the smoothed signal generated by superimposing the smoothed IMF components is shown in Figure 11(b). Figure 12 As shown.

[0103] Step D: Determine the input and target data. The input data of this invention is the time series of the sterilization period data processing area (i.e., x=1839 to x=2415), corresponding to the time data of the signals collected by FBG. The target data is the temperature signal values ​​corresponding to these time points, which are temperature signal data after decomposition and smoothing by applying the EMD algorithm.

[0104] Step E: Fit the signal.

[0105] First, create a neural network. Use computer software to create a feedforward neural network for function fitting. The hidden layer size of this neural network is 10.

[0106] Next, the dataset is divided. In this embodiment, the data processed in step C is divided into a training set (70%), a validation set (15%), and a test set (15%).

[0107] Next, the neural network is trained. This invention uses computer software to train the neural network. During training, the loss function used is the mean absolute error (MAE).

[0108] Finally, after training, the neural network is able to map new inputs (time series of sterilization period data processing area, i.e., x=1839 to x=2398) to corresponding outputs (smoothed signals).

[0109] In this embodiment, the goodness of fit is as high as 0.99934, and the fitting error is 0.1524℃. The training data and results are shown in Figure 13. In this embodiment, the network actually iterated 14 times, with an actual gradient of 0.58412, and its damping factor (Mu) has an actual value of 1 * e⁻³.

[0110] Step F: Determine the sterilization period critical point. Set the temperature threshold according to national standards to determine the start and end points. The sterilization period begins when the temperature first exceeds or equals the sterilizer's temperature threshold, and ends when the temperature last exceeds or equals the sterilizer's temperature threshold, provided that the temperature remains above the threshold throughout the period from start to end. In this embodiment, the start time is 1931 seconds (the same as the manually selected start time, 1931 seconds), and the end time is 2307 seconds (the manually selected end time is 2308 seconds, a difference of 1 second). The point determination results are for reference. Figure 14 As shown.

[0111] All the above steps can be executed and the result obtained in just 1 second.

[0112] Figures 15-20 The diagram schematically illustrates a method for determining the critical point of sterilization in a pressure steam sterilizer according to another embodiment of the present invention.

[0113] Example 3:

[0114] Taking the temperature monitoring data of the Xinhua LMQ.C-50EP high-pressure steam sterilizer (the sterilizer's temperature threshold, i.e., the sterilizer's rated operating temperature is 134℃, and the sterilization time is set to 6 minutes) as an example, the critical point of its sterilization period is automatically determined.

[0115] Step A: Signal Acquisition. A fiber Bragg grating sensor is used to acquire the raw temperature signal. In this embodiment, a total of 3525 sets of temperature data signals were acquired, containing a large amount of noise, which affects the manual selection of temperature points and leads to significant errors. (Sensing device reference) Figure 2 As shown, the original signal reference Figure 15 As shown.

[0116] Step B: Extract the optimal data processing area for determining the sterilization period. This invention uses a program to define the selection criteria for the processing area; the moment when the original signal first exceeds or equals the sterilizer temperature threshold is set as... The sterilization time set by the sterilizer is set to The optimal data processing area starting point for determining the sterilization period, as set by this invention, is... The optimal data processing region endpoint for determining the sterilization period is... Therefore, the optimal data processing area for determining the sterilization period in this invention is... In this embodiment, 30% of the sterilization time is 108 seconds, and the specific sterilization period processing area is 1965-2541 seconds. The optimal data processing area for determining the sterilization period is selected in this embodiment for reference. Figure 16 As shown.

[0117] Step C: Smooth the signal to be processed to obtain smoothed temperature data.

[0118] First, the signal obtained in step B is decomposed using the EMD algorithm to obtain eight IMF components. The IMF components obtained by the EMD algorithm are shown in Figure 17(a).

[0119] Secondly, noise removal: wavelet threshold denoising is performed on each IMF component after EMD decomposition.

[0120] In this embodiment, the threshold for:

[0121]

[0122] After wavelet threshold denoising of each IMF component, they are superimposed to obtain the signal after noise removal. The smoothed IMF components are shown in Figure 17(b), and the smoothed signal generated by superimposing the smoothed IMF components is shown in Figure 17(b). Figure 18 As shown.

[0123] Step D: Determine the input and target data. The input data of this invention is the time series of the sterilization period data processing area (i.e., x=1965 to x=2541), corresponding to the time data of the signals collected by FBG. The target data is the temperature signal values ​​corresponding to these time points, which are temperature signal data smoothed by applying the EMD algorithm.

[0124] Step E: Fit the signal.

[0125] First, create a neural network. Use computer software to create a feedforward neural network for function fitting. The hidden layer size of this neural network is 10.

[0126] Next, the dataset is divided. The data processed in step C is divided into a training set (70%), a validation set (15%), and a test set (15%).

[0127] Next, the neural network is trained. Computer software is used to train the neural network, updating the weights and biases according to the backpropagation algorithm. The loss function is the mean absolute error (MAE), which is the average of the absolute values ​​of the errors.

[0128] Finally, after training, the neural network is able to map new inputs (time series of sterilization period data processing area, i.e., x=1965 to x=2541) to corresponding outputs (smoothed signals).

[0129] In this embodiment, the goodness of fit is as high as 0.99897, and the fitting error is only 0.2017℃. The training data and results are shown in Figure 19. In this embodiment, the network actually iterated 74 times, with an actual gradient of 0.0872, and its damping factor (Mu) has an actual value of 1 * e⁻⁴.

[0130] Step F: Determine the sterilization period critical point. Set the temperature threshold according to national standards to determine the start and end points. The sterilization period begins when the temperature first exceeds or equals the sterilizer's temperature threshold, and ends when the temperature last exceeds or equals the sterilizer's temperature threshold, provided that the temperature remains above the threshold throughout the period from start to end. In this embodiment, the start time is 2046 seconds (the same as the manually selected start time, 2046 seconds), and the end time is 2429 seconds (the manually selected end time is 2434 seconds, a difference of 5 seconds). The point determination results are for reference. Figure 20 As shown.

[0131] All the above steps can be executed and the result obtained in just 1 second.

[0132] The automatic point selection algorithm of this invention can be implemented in various ways to achieve the same objective. In the basic concept of this invention, signal preprocessing, fitting, and error analysis are key steps. Therefore, in specific implementations, different preprocessing methods, fitting methods, and loss functions (i.e., error calculations) can be selected and adjusted to achieve different implementation methods.

[0133] 1. Alternatives to smoothing methods:

[0134] In this invention, the EMD method is used to smooth the signal. However, other smoothing methods can also be used, such as moving average, weighted average, median filtering, wavelet transform, etc. These methods can be selected according to actual needs and signal characteristics to achieve smoothing.

[0135] Taking the experimental data collected in Example 1 as an example, this invention uses various smoothing methods to smooth the data and compares and analyzes the effects of these smoothing methods. The main steps are as follows:

[0136] 1. Data preparation: Read the raw signal data and select the data area to be smoothed.

[0137] 2. Smoothing signal processing: Process the selected smoothing methods one by one to obtain the smoothed signal.

[0138] 3. Plotting the smoothing effect: Plot the original signal and each smoothed signal on the same graph for comparison. It can be observed that weighted average smoothing and median filtering smoothing have poor signal smoothing effect due to windowing processing, which causes signal delay.

[0139] 4. Calculation of Smoothing Effect Evaluation Index: Calculate the mean square error (MSE) between the smoothed signal and the original signal. This index helps to quantitatively evaluate the effect of the smoothing method. To better represent the smoothing effect, this invention performs a logarithmic operation on the MSE. In this embodiment, the logarithmic mean square error of each smoothing method is shown below:

[0140]

[0141] 5. Smoothing Effect Comparison and Analysis: Based on the calculated evaluation indicators, the effects of various smoothing methods are compared and analyzed. This invention uses bar charts for visualization.

[0142] 6. Results Interpretation and Selection of the Best Method: Based on the comparative analysis results, interpret the differences in effectiveness among various smoothing methods and select the best smoothing method. Methods with smaller mean squared error or other superior evaluation metrics can be considered as the optimal choice.

[0143] 7. Results Verification and Adjustment: The selected optimal smoothing method was verified, and it was found that EMD smoothing treatment met expectations and achieved the best results in practical applications.

[0144] Through the above process and approach, we can analyze and compare smoothing methods, and select the EMD smoothing method as the optimal one, followed by the wavelet transform smoothing method, both of which can meet the needs of signal processing.

[0145] Reference for smoothing effects of various smoothing methods Figure 21 As shown, a comparison of the smoothing effects of each method is provided for reference. Figure 22 As shown, the smoothing effect evaluation chart is for reference. Figure 23 As shown.

[0146] columnar Figure 23 This method is used to evaluate the effectiveness of different smoothing methods in setting the critical time point during sterilization. The MSE value represents the average of the squared differences between the smoothed signal and the original signal; a smaller MSE value indicates better smoothing. Based on the results of the bar chart, it can be found that the EMD smoothing method is the most effective for smoothing temperature signals in pressure steam sterilizers, followed by the wavelet transform smoothing method.

[0147] 2. Alternatives to the fitting method:

[0148] This invention uses a neural network for signal fitting. However, other fitting methods can also be used, such as polynomial fitting, spline interpolation, and radial basis functions. These methods can be selected according to actual needs and signal characteristics to achieve the desired fitting.

[0149] Neural network fitting mainly relies on the following steps and algorithm principles:

[0150] 1. Define the neural network structure: A neural network consists of multiple layers of neurons (nodes), including an input layer, hidden layers, and an output layer. The input layer receives data, the hidden layers process the data, and the output layer outputs the results. The number of neurons in each layer is set according to the complexity of the problem and the requirements.

[0151] 2. Initialize parameters: The parameters (weights and biases) of a neural network are usually initialized to random values. This helps to break the symmetry and allows each hidden unit to learn different features.

[0152] 3. Forward Propagation: In the forward propagation phase, the neural network calculates the prediction result based on the input and the current parameters. This process starts from the input layer and proceeds layer by layer until the output layer.

[0153] 4. Loss Calculation: The loss function measures the difference between the network prediction and the true label. In this invention, the loss function is calculated using the mean absolute error.

[0154] 5. Backpropagation: Backpropagation is an efficient method for calculating the gradient of the loss function with respect to the network parameters. During backpropagation, the error is propagated backward from the output layer to the input layer to update the network's weights and biases.

[0155] 6. Update Weights: The neural network uses gradient descent (or its variants, such as stochastic gradient descent, Adam, etc.) to update its weights and biases based on the gradient of the loss function. This step aims to reduce prediction error and optimize network performance.

[0156] 7. Iterative training: Repeat the steps of forward propagation, loss calculation, backpropagation, and weight update until the model performs satisfactorily on the training set or reaches the preset number of iterations.

[0157] 8. Evaluation and Prediction: After the model is trained, its performance needs to be evaluated on an independent test dataset.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A time-point method for the critical point of the sterilization cycle of a pressure steam sterilizer, wherein, The method comprises the following steps: A, signal acquisition: obtaining the temperature original signal of the pressure steam sterilizer through the sensor; B, intercepting the optimal data processing area of the original signal for judging the sterilization period to obtain the to-be-processed signal; C, decomposing the to-be-processed signal obtained in step B by using the EMD algorithm to obtain a plurality of IMF components, then performing wavelet threshold denoising on the IMF components, and superimposing the denoised IMF components to obtain the temperature data after smoothing processing; D, determining input data and target data: the input data is the time sequence of the optimal data processing area for judging the sterilization period, and the time data corresponding to the signal collected by the sensor; The target data is the temperature data corresponding to the time data after being processed by step C; E, signal fitting: mapping the input data to the temperature data after being processed by step D, calculating the mean absolute error between the fitting signal and the target signal to quantify the fitting error, and creating a feedforward neural network for function fitting; Divide the data set, divide the data processed in step C into a training set, a validation set and a test set, and train the neural network; F, determining the critical point of the sterilization period: the starting point of the sterilization period is the point at which the temperature is greater than or equal to the temperature threshold of the sterilizer for the first time, and the ending point is the point at which the temperature is greater than or equal to the temperature threshold of the sterilizer for the last time, and the temperature during the starting point to the ending point is greater than the temperature threshold; The sensor is a Bragg fiber grating sensor.

2. The method of claim 1, wherein, In step B, the first time point in the original signal greater than or equal to the sterilizer temperature threshold value is set as T, the sterilization time set by the sterilizer is set as t, and the start of the optimal data processing region for judging the sterilization period is set as T B =T-t*30%, the end of the optimal data processing region for judging the sterilization period is set as T E =T B +t*(1+30%), and the optimal data processing region for judging the sterilization period is [T B ,T E ].

3. The method of claim 1, wherein, In step C, the to-be-processed signal is smoothed by using moving average, weighted average, median filtering or wavelet transform.

4. The method of claim 1, wherein, The size of the hidden layer of the neural network is 10.

5. The method according to any one of claims 1 to 4, wherein, 70% of the temperature data after smoothing processing in step C is divided into the training set, 15% is divided into the validation set, and 15% is divided into the test set.

Images