A method and system for rapid detection of water activity

Abstract

The application belongs to the technical field of detection, and particularly relates to a rapid detection method and system for water activity, which comprises the following steps: using a water activity sensor or a detection device to realize real-time sensing and measurement of water activity values of a sample at different time points; correlatively storing time stamps and water activity values of obtained real-time water activity data; performing denoising, smoothing and formatting processing on collected original data, and performing data caching according to a set time interval or a real-time updating mode; obtaining a fitting curve by real-time fitting of a water activity time sequence curve according to discrete water activity data and based on a water activity prediction model, and deducing a water activity value at a future time according to the fitting curve; and performing difference comparison according to two continuous fitting curves to obtain a difference sequence. The application can obtain a final stable value of water activity quickly under the premise of a small amount of initial detection data, through algorithm fitting prediction, combined with continuous data updating and stability judgment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of detection, and particularly relates to a rapid detection method and system for water activity. BACKGROUND

[0002] Water activity is an important indicator for measuring the content of free water in a substance, and is widely used in food, medicine, cosmetics, chemical industry and other industries. Water activity not only affects the physical properties, chemical stability and microbial growth of products, but also directly relates to the quality, safety and shelf life of products. Therefore, accurate and rapid determination of water activity is of great significance for production process control, product research and development, and quality management.

[0003] At present, the determination of water activity mainly relies on traditional physical measurement methods. The measurement process involves sample balancing, equipment calibration and data recording steps, and the entire determination process often takes a long time, usually several hours or even several days to obtain stable measurement results. This long determination cycle is particularly inconvenient in fast production and high real-time quality control application scenarios, limiting its application in real-time monitoring and rapid response.

[0004] In addition, the traditional water activity determination method has high requirements for environmental conditions during operation, such as improper control of temperature and humidity, which may lead to inaccurate or unstable measurement results. This not only increases the uncertainty of measurement, but also may affect the continuity and efficiency of the production process.

[0005] For production processes that require rapid feedback and real-time quality control, low determination efficiency not only reduces production efficiency, but also makes it difficult to timely discover and correct product quality problems. Therefore, there is an urgent need for a detection method that can quickly predict the final stable value of water activity at the initial stage of determination, so as to greatly shorten the test time and improve the real-time monitoring capability of the production process. SUMMARY

[0006] The purpose of the present application is to provide a rapid detection method and system for water activity, which can obtain a small amount of determination data at the initial stage, predict by algorithm fitting, and quickly obtain the final stable value of water activity by combining continuous data updating and stability judgment.

[0007] The technical solutions adopted by the present application are as follows:

[0008] A rapid detection method and system for water activity, comprising:

[0009] using a water activity sensor or a determination device to sense and measure the water activity value of the sample at different time points in real time;

[0010] storing the time stamp corresponding to the real-time water activity data and the water activity value in association;

[0011] The collected raw data is denoised, smoothed, and formatted, and is cached according to a set time interval or a real-time update mode;

[0012] According to the discrete water activity data, a fitting curve is obtained by real-time fitting of the water activity time series curve based on a water activity prediction model, and a water activity value at a future time is inferred according to the fitting curve;

[0013] The difference between two consecutive fitting curves is compared to obtain a difference sequence, and a statistical analysis is performed based on the difference sequence to obtain a determination result of the stability of the fitting curve;

[0014] The stability determination result is combined with the latest fitting curve to output the final water activity measurement value and related data.

[0015] As a preferred scheme, the measurement device includes water activity measurement using a dew point method, a resistance method, or a microwave absorption method.

[0016] As a preferred scheme, the dew point method specifically includes the following steps:

[0017] The solid sample is crushed to a particle size of not greater than 5 mm, the liquid or semi-liquid sample is uniformly stirred, and the sample filling amount is controlled to be two-thirds of the sample cup capacity;

[0018] The sample cup is quickly placed in the measurement cabin, and after being closed, the mirror surface temperature is rapidly reduced from the ambient temperature to below the expected dew point under a constant temperature and humidity environment through a semiconductor module, and then is raised at a rate of 0.1℃ / second, and the mirror surface light reflectivity is scanned in real time through an infrared laser to capture the dewing moment;

[0019] The system automatically records the dewing critical temperature, and simultaneously calculates the water activity based on the critical temperature according to the Clausius-Clapeyron equation.

[0020] As a preferred scheme, the resistance method specifically includes the following steps:

[0021] The solid sample is crushed to a particle size of not greater than 3 mm, the liquid or semi-liquid sample is uniformly stirred, and the sample filling amount is controlled to be two-thirds of the sample cup capacity;

[0022] The sample cup is quickly placed in the measurement cabin, and after being closed, the sample is absorbed by a hygroscopic material under a constant temperature and humidity environment, and the real-time resistance value is measured using a sensor, and the steady-state resistance value is recorded;

[0023] The steady-state resistance value is converted to a water activity value using an instrument built-in algorithm based on the steady-state resistance value.

[0024] As a preferred scheme, the microwave absorption method specifically includes the following steps:

[0025] Solid samples should be crushed to a particle size of no more than 3mm. Liquid or semi-fluid samples should be stirred evenly, and the sample filling thickness should be controlled within 2mm-5mm.

[0026] The sample is placed in a sample dish made of microwave-transmitting material. Microwaves of a fixed frequency are emitted, penetrate the sample, and then the attenuation signal is received. The energy loss value after the microwave passes through the sample is recorded by the sensor.

[0027] Based on the energy loss value and the correlation between the absorption intensity of microwaves by water and molecular polarity, a mathematical model of loss factor and water activity was established to calculate the water activity.

[0028] As a preferred solution, the specific steps of the associated storage are as follows:

[0029] Data reception and labeling: Collect and receive water activity measurement values ​​and record the timestamp of the measurement moment. Each data point is composed of a timestamp and a water activity value so that its time series characteristics can be identified later.

[0030] Data storage or caching: The received data points are stored in the system cache or database in chronological order. At the same time, the system is networked and transmits the data to the host computer or cloud for backup through the communication interface;

[0031] Abnormal data filtering: Mark or remove data whose values ​​are outside the reasonable range; or use historical data comparison or statistical methods to identify abnormal points.

[0032] As a preferred solution, the specific method for denoising, smoothing and formatting the raw data includes a moving average method or a Kalman filter method;

[0033] in,

[0034] The moving average method specifically comprises the following steps:

[0035] Determine the window size and weight distribution rules based on data characteristics;

[0036] Arrange the data in ascending order according to the time series by timestamp. Missing values ​​need to be interpolated or removed, and the moving average is calculated after removing obvious outliers.

[0037] The Kalman filtering method specifically comprises the following steps:

[0038] Define the initial state vector and its error covariance matrix according to the data type. If the initial state is unknown, it can be set as a diagonal matrix;

[0039] Predict the state at the next moment based on the system dynamic model and update the uncertainty of the predicted state;

[0040] The weight coefficients of the prediction and observation are fused, and the actual observation value is used to correct the predicted state, and the corrected state uncertainty is updated;

[0041] The predicted state and state uncertainty are taken as the input of the next time, and the prediction and correction process is repeated to realize real-time recursive filtering.

[0042] As a preferred scheme, the specific fitting step in the water activity prediction model is as follows:

[0043] Preliminary fitting: after obtaining a small amount of discrete data points at the beginning of the determination, a first fitting curve is generated;

[0044] Dynamic updating: as the detection proceeds, new data points are continuously added, and the existing water activity prediction model is updated or re-fitted;

[0045] Real-time output of prediction results: based on the latest fitting curve, the water activity prediction value after the current time point is obtained.

[0046] As a preferred scheme, the specific output step of the determination result is as follows:

[0047] Compare the continuous fitting curves: after obtaining two continuous fitting curves, select the water activity values y corresponding to the same time points x and compare them to obtain a difference sequence;

[0048] Calculate the difference statistics: statistically analyze the difference sequence, specifically calculate the standard deviation or maximum difference;

[0049] Threshold determination: when the difference statistics is lower than the preset threshold, it is determined that the fitting curve has tended to be stable, indicating that the final value of water activity no longer changes significantly;

[0050] Output results: once the stability is determined, a "determination stable" signal is sent to the result output module for final result extraction and presentation.

[0051] A rapid detection system for water activity, which applies the rapid detection method for water activity described above, comprising:

[0052] A sensing and detection module, including a water activity sensor or a determination device, for real-time sensing and measuring the water activity values of the sample at different time points;

[0053] A data acquisition module for acquiring real-time water activity data from the sensing and detection module, and storing the corresponding time stamp and water activity value in association;

[0054] A data processing module for denoising, smoothing, and formatting the original data collected by the data acquisition module, and performing data caching according to the set time interval or real-time updating mode;

[0055] a data fitting and prediction module, configured to perform real-time fitting on the water activity time curve according to the discrete water activity data output by the data processing module by using a water activity prediction model to obtain a fitting curve, and deduce the water activity value at a future time according to the fitting curve;

[0056] a stability judgment module, configured to perform difference comparison on two consecutive fitting curves to obtain a difference sequence, and perform statistical analysis based on the difference sequence to obtain a judgment result of the stability of the fitting curve;

[0057] a result output module, configured to combine the stability judgment result with the latest fitting curve to output a final water activity measurement value and related data.

[0058] The technical effects achieved by the present application are as follows:

[0059] The present application can quickly fit and continuously update the initial small amount of data by using the water activity prediction model, without waiting for a long time for the sample and the measurement environment to be completely balanced, so that accurate water activity prediction values can be obtained at an early stage. Compared with the prior art, the measurement efficiency is significantly improved. Meanwhile, in the field where water activity is closely related to product quality, the present application can quickly discover abnormality or trend change, so that corresponding measures can be taken in time to improve the overall quality control and safety management level.

[0060] The present application introduces "continuous updating" and "stability judgment" into the detection process to form an indispensable new step. This step allows the system to continuously correct the fitting model after obtaining new data in real time, and to make a judgment in time when the fitting curve shows signs of stability, so as to ensure the accuracy and stability of the detection result.

[0061] The present application shortens the measurement period, so that the production line can quickly obtain water activity related information, thereby more flexibly adjusting and reducing the loss or rework caused by deviation of water activity from the target value. At the same time, the occupancy time of the detection equipment is reduced, and the operating cost is also reduced.

[0062] The method of the present application does not depend on a specific type of sensor, but only requires corresponding accurate measurement and data communication capabilities. It is suitable for various scenes in the food, pharmaceutical, cosmetic, chemical and other industries, has low requirements for production condition modification, and has good versatility and flexibility. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is a flow structure schematic diagram of the rapid detection method of water activity in the present application;

[0064] Figure 2 is a flow structure schematic diagram of the rapid detection system of water activity in the present application. DETAILED DESCRIPTION

[0065] In order to make the purpose and advantages of the present invention more clearly understood, the present invention is described in detail below with reference to the following examples. It should be understood that the following text is only used to describe one or more specific embodiments of the present invention and does not strictly limit the scope of protection of the present invention.

[0066] like Figure 1 As shown in the figure, a rapid detection method for water activity is proposed. Its core idea is: on the premise of obtaining a small amount of measurement data in the initial stage, fitting prediction is performed through an algorithm, and combined with continuous data updates and stability judgment, the final stable value of water activity is quickly obtained.

[0067] The specific steps are as follows:

[0068] Step 1: Using a water activity sensor or measuring device to sense and measure the water activity value of the sample at different time points in real time;

[0069] Step 2: Associate and store the corresponding timestamp of the acquired real-time water activity data with the water activity value;

[0070] Step 3: De-noise, smooth, and format the collected raw data, and cache the data according to the set time interval or real-time update mode;

[0071] Step 4: Based on the discrete water activity data and the water activity prediction model, the water activity time series curve is fitted in real time to obtain a fitting curve, and the water activity value at a future moment is inferred based on the fitting curve;

[0072] Step 5: Compare the difference between the two consecutive fitting curves to obtain a difference sequence, and perform statistical analysis based on the difference sequence to obtain a determination result of the stability of the fitting curve;

[0073] Step 6: Combine the stability judgment result with the latest fitting curve to output the final water activity measurement value and related data.

[0074] like Figures 1-2 As shown, a rapid water activity detection system is provided, to which the above-mentioned rapid water activity detection method is applied. The system specifically includes the following functional modules and steps:

[0075] 1. Sensing Detection Module

[0076] Function and effect: This module mainly includes a water activity sensor or measuring device, which is used to sense and measure the water activity value of the sample at different time points in real time.

[0077] Implementation: The dew point method, resistance method, microwave absorption method, and other existing water activity measurement means can be used, or other suitable hardware devices can be selected according to specific applications. The module needs to ensure that the measurement environment is relatively stable in terms of temperature, humidity, and other conditions, reducing measurement noise and errors.

[0078] It should be noted that different water activity sensors or measurement devices can be selected according to different experimental types and product types. In this embodiment, the Decagon Aqualab 4TE water activity sensor is selected. Of course, the specific steps for water activity measurement using the dew point method, resistance method, and microwave absorption method are as follows:

[0079] First, the dew point method specifically includes the following steps:

[0080] 1. Sample pretreatment

[0081] 1.1 Sample preparation

[0082] Solid samples need to be crushed to a particle size of ≤5 mm, and liquid samples need to be uniformly stirred to ensure uniform moisture distribution.

[0083] The sample filling amount is controlled at 2 / 3 of the sample cup capacity to avoid excessive compression affecting vapor equilibrium.

[0084] 1.2 Environmental calibration

[0085] The laboratory needs to be constant temperature (25±0.5℃) and humidity 50%-80%, and the instrument needs to be preheated for 30 minutes to eliminate temperature drift.

[0086] 2. Start the closed balance system

[0087] Seal the sample cabin: quickly place the sample cup into the measurement cabin to ensure good air tightness after the cabin door is closed, preventing environmental humidity interference.

[0088] Note: If the sample contains volatile components (such as alcohol), a special sealing film needs to be used for blocking.

[0089] 3. Mirror refrigeration and dew point detection

[0090] 3.1 Dynamic cooling control mirror

[0091] Through the semiconductor module, the mirror temperature is rapidly reduced from the ambient temperature to below the expected dew point (usually -10℃), and then it is raised at a rate of 0.1℃ / second.

[0092] The infrared laser scans the mirror reflectivity in real time, capturing the dewing moment (reflectivity drop ≥15%).

[0093] 3.2 Dew point temperature locking

[0094] The system automatically records the dewing critical temperature Td (dew point temperature) with an accuracy of ±0.1℃.

[0095] 4. Water activity calculation

[0096] 4.1 Vapor pressure conversion

[0097] Water activity calculation according to Clausius-Clapeyron equation :

[0098] ;

[0099] Where, a is water activity, P is sample vapor pressure, Psat is pure water saturated vapor pressure, L is latent heat of vaporization, R is water vapor gas constant, Tref is reference temperature, Td is dew point temperature.

[0100] Secondly, the resistance method specifically includes the following steps:

[0101] 1. Instrument calibration (core pretreatment)

[0102] 1.1 Standard salt solution calibration

[0103] Use saturated salt solution (such as sodium chloride aw=0.753 or magnesium chloride aw=0.328) to inject the calibration cup, covering the sensor surface.

[0104] Set the calibration mode according to the instrument manual, wait for the reading to stabilize (usually 10-15 minutes), and the error should be ≤±0.01 aw. It should be noted that the calibration frequency is once a week, or immediately after the environmental temperature and humidity fluctuate sharply.

[0105] 2. Sample preparation

[0106] 2.1 Homogenization treatment

[0107] Solid samples are crushed to a particle size of ≤3mm (such as milk powder, drug coatings), to avoid the influence of interstitial gaps on vapor equilibrium.

[0108] Liquid / semi-fluid samples (such as sauces) need to be thoroughly stirred to eliminate stratification.

[0109] 2.2 Sample cup filling

[0110] Fill the volume to 2 / 3 of the cup capacity, ensuring full contact with the sensor, and excessive compaction may cause vapor diffusion to be blocked.

[0111] 3. Sensor loading and equilibration

[0112] 3.1 Closed chamber operation

[0113] Put the sample cup into the measuring chamber, seal it, and start the air circulation (flow rate 1-2 L / min) after the humidity balance in the chamber is accelerated.

[0114] 3.2 Temperature balance control

[0115] The sample temperature needs to be consistent with the sensor temperature (temperature difference ≤0.5℃), and the balance time is usually 20 minutes to 2 hours.

[0116] It is worth noting that high oil / glycerin samples need to extend the balance time to more than 1.5 hours.

[0117] 4. Resistance measurement and data acquisition

[0118] 4.1 Electrical signal conversion

[0119] The resistance value of the sensor hygroscopic material (such as lithium chloride) decreases after hygroscopic, and the instrument records the steady-state resistance value.

[0120] 4.2 Water activity conversion

[0121] The instrument built-in algorithm converts the resistance value to value, the formula is as follows:

[0122] ;

[0123] Among them, is the water activity value, is the measured resistance, , is the calibration curve parameter.

[0124] Thirdly, the microwave absorption method specifically includes the following steps:

[0125] 1. Instrument calibration and benchmark establishment

[0126] 1.1 Standard substance calibration

[0127] Use standard salt solution (such as sodium chloride aw=0.753) with known water activity or standard humidity sheet to fill the sample cavity.

[0128] Record the corresponding relationship between microwave energy absorption rate and standard value, and construct the calibration curve (usually 3-5 standard points are needed).

[0129] 2. Sample preparation and loading

[0130] 2.1 Sample pretreatment

[0131] Solid samples need to be crushed to a particle size of ≤3mm, and liquid samples need to be uniformly stirred to avoid stratification affecting microwave penetration.

[0132] The sample thickness is controlled at 2mm-5mm (too thick will cause signal attenuation, and too thin will not be sensitive enough).

[0133] 2.2 Sample chamber loading

[0134] Place the sample in a dedicated sample dish made of microwave-transparent material (e.g. Teflon), ensuring a flat surface.

[0135] 3. Microwave parameter setting and measurement

[0136] 3.1 Microwave transmission and reception

[0137] Transmit a fixed-frequency microwave (common frequency bands: 2.45 GHz or 5.8 GHz), and receive the attenuated signal after penetrating the sample.

[0138] Control the microwave power within the range of 10 mW-50 mW to avoid thermal effects interfering with the measurement.

[0139] 3.2 Real-time monitoring of energy loss

[0140] Record the energy loss value of the microwave after passing through the sample through the sensor (unit: dB), the calculation formula is:

[0141] ;

[0142] Where, is the energy loss value, is the incident power, is the transmitted power.

[0143] 4. Calculation of water activity

[0144] 4.1 Loss factor correlation water activity

[0145] The absorption intensity of water on microwaves is related to the molecular polarity, and a mathematical model of the loss factor and is established:

[0146] ;

[0147] Where, is the water activity value, is the energy loss value, , are the calibration curve parameters.

[0148] Output result: send the real-time measured water activity value (and the corresponding timestamp) to the data acquisition module.

[0149] II. Data acquisition module

[0150] Function and role: This module is responsible for obtaining real-time water activity data from the sensor detection module, and associating and storing the corresponding timestamp and water activity value.​

[0151] Implementation: Data acquisition cards or built-in communication interfaces can be used to communicate with sensors through serial ports, USB, network interfaces, etc. At the same time, preliminary filtering or anomaly detection (such as removing obviously distorted data) is performed on the data.

[0152] Output results: Discrete data points of time-water activity for subsequent data processing and model fitting.

[0153] Three, data processing module

[0154] Function and role: This module performs denoising, smoothing, and formatting on the collected raw data, and performs data caching according to the set time interval or real-time update mode.

[0155] Implementation: According to experience or specific application requirements, methods such as moving average, Kalman filtering, etc. can be used to process data to reduce random errors and maintain data authenticity.

[0156] It is worth noting that when using the moving average method to process data, the window size and weight distribution rules need to be determined according to the data characteristics; then arrange the data in ascending order according to the time stamp, and interpolate or remove missing values, and calculate the moving average after removing obvious outliers.

[0157] Secondly, when using Kalman filtering, the initial state vector and its error covariance matrix need to be defined according to the data type, and if the initial state is unknown, it can be set as a diagonal matrix; then predict the next state according to the system dynamic model, and update the uncertainty of the predicted state; then fuse the weight coefficients of prediction and observation, and use the actual observation value to correct the predicted state, and update the uncertainty of the corrected state; finally, the predicted state and state uncertainty are used as the input of the next time, and the prediction and update process is repeated to realize real-time recursive filtering.

[0158] Output results: Generate high-quality, structured time-water activity data stream to provide reliable data input for fitting and prediction modules.

[0159] Four, data fitting and prediction module

[0160] Function and role: This module is the core of the invention. Its main purpose is to use mathematical models (such as nonlinear least squares, polynomial fitting, logarithmic function fitting, or machine learning-based prediction models) to fit the water activity time series curve in real time based on the discrete water activity data output by the data processing module, and to infer the water activity value at future time.

[0161] Implementation:

[0162] Initial fitting: After obtaining a small number of discrete data points at the beginning of the measurement, the first fitting curve is generated;

[0163] Dynamic updating: As the detection progresses, new data points are continuously added, and the existing model is updated or refitted;

[0164] Real-time output of prediction results: Based on the latest fitting curve, the water activity prediction value after the current time point is obtained.

[0165] Output results: Real-time updated fitting curve and prediction value provide input basis for stability judgment module.

[0166] Five, stability judgment module

[0167] Function and role: Used to determine whether the water activity prediction result has stabilized, avoiding the defect that the final result can only be confirmed after a long wait in traditional methods.

[0168] Implementation:

[0169] Compare the continuous fitting curves: After obtaining two consecutive fitting curves, select the water activity values (y) at the same time points (x) for comparison, and obtain the difference sequence;

[0170] Calculate the difference statistics: Perform statistical analysis on the difference sequence, typically by calculating the standard deviation or maximum difference;

[0171] Threshold determination: When the difference statistics is lower than the preset threshold (set in advance according to different experimental scenarios), it is determined that the fitting curve has stabilized, indicating that the final value of water activity no longer changes significantly.

[0172] Output results: Once it is determined to be stable, send a "measurement stable" signal to the result output module for final result extraction and presentation.

[0173] Six, result output module

[0174] Function and role: Combine the stability judgment result with the latest fitting curve to output the final water activity measurement value and related data.

[0175] Implementation: The final predicted water activity value can be displayed on the instrument panel or software interface, and a measurement report (including measurement time, sensor information, prediction model type, stability determination curve, etc.) can be recorded and generated.

[0176] Output results: Provide the final water activity value to the user or production control system for subsequent quality control or process management.

[0177] As Figures 1-2 shown, a specific application of a rapid water activity detection method and system, its main application fields are as follows:

[0178] 1. Food Industry

[0179] Raw Material and Recipe Evaluation: For coffee beans, grains, seasonings, baking materials, dairy products, meat products, and aquatic products, controlling water activity can effectively prevent microbial growth and ensure flavor and texture.

[0180] Production Process Control: Monitor water activity in real-time during food processing, adjust drying, moisturizing, or sterilization processes in a timely manner, and improve product consistency and production efficiency.

[0181] Storage and Shelf Life Research: By measuring water activity, evaluate the quality changes of food under different storage conditions, optimize packaging and storage methods, and extend shelf life.

[0182] 2. Pharmaceutical and Biotechnology Industry

[0183] Drug Development and Quality Control: For tablets, capsules, injection powders, and freeze-dried preparations, precise control of water activity helps avoid drug failure, deterioration, or microbial contamination.

[0184] Production of Strains and Enzyme Preparations: In microbial fermentation and enzyme preparation production, monitor water activity to ensure the stability and effectiveness of bioactive substances.

[0185] 3. Cosmetics and Personal Care Products Industry

[0186] Product Stability Evaluation: For creams, masks, and cleaning and care products, water activity has a significant impact on microbial growth and product texture. Monitor and control water activity to extend shelf life and maintain product efficacy.

[0187] New Product Development: Monitor water activity during formula design to optimize formulations to ensure product stability and safety in different environments.

[0188] 4. Research Institutions and University Laboratories

[0189] Basic Research: In food science, pharmacy, microbiology, and materials science research experiments, water activity is an important factor affecting various reactions and processes. Precise measurement helps reveal the interaction mechanism between water and materials or microorganisms.

[0190] 5. Other Potential Fields

[0191] Electronic Device and Lithium Battery Manufacturing: Some high-end electronic components and lithium battery electrolytes have strict humidity requirements. Water activity monitoring can prevent quality or safety hazards.

[0192] In the application process in the above fields, the specific implementation methods are as follows:

[0193] Step One: Sensing and Detection Module

[0194] 1.1 Sample Preparation and Placement

[0195] Place the sample to be tested into the sample chamber or appropriate measurement location of the water activity detector, ensuring that the basic conditions required by the instrument (such as temperature, humidity, etc.) meet or approach the detection requirements.

[0196] Check whether the sensor is calibrated and whether the equipment meets the measurement standards.

[0197] 1.2 Start Sensor Measurement

[0198] Start the water activity sensor for data collection through the water activity detector or control system interface.

[0199] The sensor can use continuous collection mode or interval collection mode (e.g., measuring once every few seconds / minutes), and the specific frequency can be configured according to actual needs.

[0200] 1.3 Measure Water Activity Value

[0201] At each sampling moment, the sensor will immediately measure the water activity value of the sample and temporarily store the measurement value in the buffer area inside the detector or directly output it to the next module.

[0202] During this process, it is necessary to ensure that the measurement environment is relatively stable to minimize data fluctuations or noise caused by external interference.

[0203] Operation Points

[0204] Ensure that the sensor sensitivity meets the standard. If it is a dew point sensor, a resistance sensor, or a microwave absorption sensor, it should be calibrated in advance.

[0205] The sampling frequency should be matched with the expected detection speed; higher sampling frequency can capture water activity changes faster, but may bring more noise or increase system load.

[0206] Step Two: Data Collection Module

[0207] 2.1 Data Reception and Labeling

[0208] The data collection module receives water activity measurement values from the sensing and detection module and records the time stamp of the measurement time.

[0209] Each data point is composed of (time stamp, water activity value) so that the subsequent module can identify its time sequence characteristics.

[0210] 2.2 Data Storage or Buffering

[0211] Store the received data points in time order in the system buffer or database.

[0212] System networking through communication interface (such as serial, USB, network interface, etc.) to the host computer or cloud backup.

[0213] Abnormal data filtering

[0214] In order to avoid the subsequent analysis process is obviously wrong data interference, can be added in this step of preliminary abnormal value judgment, such as the data of the value beyond the reasonable range of marking or elimination.

[0215] The operation can be designed as a simple threshold judgment, but also use historical data comparison or statistical method to identify abnormal points.

[0216] Operation points

[0217] The synchronization and timing accuracy of data acquisition is very important for subsequent fitting and analysis.

[0218] The acquisition frequency should be consistent with the sensor detection module, and the system delay should be reduced as much as possible.

[0219] Step three: data processing module

[0220] 3.1 De-noising

[0221] The original data collected is filtered or smoothed, and the common methods include moving average filter, Kalman filter or wavelet denoising, etc., the purpose is to remove high frequency noise.

[0222] This step can significantly improve the usability of the original data, and avoid the influence of excessive data jitter on subsequent fitting.

[0223] 3.2 Abnormal detection and correction

[0224] For discrete data points outside the normal fluctuation range of water activity value, determine again. If it is confirmed that the data point is abnormal, it can be eliminated or corrected by interpolation.

[0225] In addition, if the sensor is disturbed during measurement (such as sudden vibration, sample contamination, etc.), repeated measurement or reference to auxiliary sensor data should be used for correction.

[0226] 3.2 Data formatting

[0227] After the above de-noising and abnormal detection, the relative accurate and stable water activity time series data is obtained.

[0228] These data are packaged or cached as "processed time series data stream" with specific structure (such as two-dimensional array or key-value pair) for the next fitting and prediction module.

[0229] Operation points

[0230] The selection of filtering method should consider real-time and smoothness, and avoid excessive filtering leading to loss of effective information.

[0231] This step can be considered as a key link to improve data quality and prevent invalid data from affecting the algorithm.

[0232] Step four: fitting and prediction module

[0233] 4.1 Preliminary fitting

[0234] Obtain the initial data that has been processed (usually the data within several or dozens of minutes at the beginning of the test), and use the selected mathematical model for preliminary fitting.

[0235] Obtain the processed water activity data, construct the initial fitting model, and use the following formula:

[0236] ;

[0237] Where:

[0238] : humidity value;

[0239] : time;

[0240] : slope;

[0241] : intercept.

[0242] The fitting formula estimates the parameters using the following method:

[0243] ;

[0244] Where:

[0245] : design matrix (contains the logarithm of time and constant term);

[0246] : weight matrix, updated by BISQUARE method;

[0247] : logarithmic value of humidity.

[0248] The power function model obtained by fitting is:

[0249] .

[0250] 4.2 Continuous data updating and dynamic fitting

[0251] As time goes on, new water activity data will be continuously sent into this module. At this time, the original fitting model can be modified or retrained using methods such as Rolling Update or Incremental Learning. The addition of new data allows the model to gradually correct the bias caused by initial estimation and better fit the actual water activity trend.

[0252] 4.3 Real-time prediction output

[0253] After obtaining the latest fitting curve, the fitting will be This module can give the current time and future short-term water activity prediction values, which are provided to the next module for stability judgment or directly used for display.

[0254] When the measurement time is still in the early stage, the prediction result may have some error, but as new data is continuously injected, the prediction accuracy will improve.

[0255] Operation points

[0256] Different fitting models have different applicability to data size and change characteristics. Logarithmic curve often has good fitting degree for many water activity equilibrium over time scenarios.

[0257] The update strategy of dynamic fitting needs to be designed according to the sampling frequency, model complexity, etc. It needs to ensure the prediction accuracy and also consider the operation efficiency.

[0258] Step five: stability judgment module

[0259] 5.1 Compare consecutive fitting curves

[0260] After updating the fitting curve each time, compare the new curve with the previous (or previous) fitting curve.

[0261] Let the prediction value of the nth fitting curve be , and the prediction value of the (n-1)th fitting curve be , then the difference between the two curves at a certain time point is:

[0262] ;

[0263] The difference sequence at all time points is:

[0264] ;

[0265] Where is the total number of time points.

[0266] 5.2 Calculate difference statistics

[0267] Statistical analysis is performed on the resulting difference series.

[0268] The mean square error is used to measure the overall deviation between two fitted curves:

[0269] ;

[0270] The standard deviation measures the range of distribution of the prediction differences:

[0271] ;

[0272] Where:

[0273] is the mean value of the difference series, calculated as:

[0274] ;

[0275] The maximum deviation is the largest absolute value in the difference series:

[0276] ;

[0277] 5.3 Stability judgment

[0278] When the difference statistics meet the threshold requirements:

[0279] and ;

[0280] Where:

[0281] : Mean square error threshold;

[0282] : Maximum deviation threshold.

[0283] It is determined that the current fitted curve has basically reflected the true water activity balance trend. At this time, the variation range of the prediction value is small, and the addition of new data will only bring slight correction.

[0284] If the statistics are still significantly high, it means that the prediction curve is still in the convergence process, and more data needs to be collected and the "fitting and prediction" step needs to be repeated.

[0285] Operation points

[0286] Selecting appropriate threshold , is the key, which can be determined based on experience or through pre-experiments.

[0287] Step 6: Final result output module

[0288] 6.1 Extract stable water activity value

[0289] Once the fitting curve is determined to be stable, the predicted water activity value at the current time or theoretically at the next time can be read from the curve as the final determination result.

[0290] Ideally, the water activity value will be in the flat or asymptotic region of the curve, and close to the traditional long-time equilibrium measurement result.

[0291] 6.2 Data display and interaction

[0292] On the display screen of the detector or the interface of the upper system, the current water activity value, the current fitting curve, and the stability determination state are displayed in real time.

[0293] If in the intelligent production environment, the result can be pushed to the factory execution system (MES) or quality management system (QMS), which is convenient for making production control decisions quickly.

[0294] 6.3 Record and archive

[0295] The final detection result, fitting model parameters, stability determination time, and other related information are recorded in the detector or the background database, which is convenient for subsequent traceability, analysis, or audit. If necessary, data can be continuously collected and log monitoring can be performed, and long-term tracking of abnormal conditions or new formulations / new products can be performed.

[0296] 6.4 End

[0297] Complete the rapid determination process of the current batch or the current sample. If a new sample needs to be determined, the process can be reset or the next detection can be restarted.

[0298] Operation points

[0299] For occasions with extremely high accuracy requirements, it is recommended to perform a short-term supplementary measurement based on the predicted result to ensure that there is no mistake.

[0300] The above is only the preferred embodiment of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the scope of protection of the present application. The structures, devices, and operation methods not specifically described and explained in the present application, such as without special description and limitation, are implemented according to the conventional means in the art.

Claims

1. A rapid detection method for water activity, characterized in that: include: Using a water activity sensor or measuring device to sense and measure the water activity value of the sample at different time points in real time; The corresponding timestamp of the acquired real-time water activity data is associated with the water activity value and stored; De-noise, smooth, and format the collected raw data, and cache the data according to the set time interval or real-time update mode; According to the discrete water activity data and based on the water activity prediction model, a water activity time series curve is fitted in real time to obtain a fitting curve, and the water activity value at a future moment is inferred based on the fitting curve; Comparing the difference between two consecutive fitting curves to obtain a difference sequence, and performing statistical analysis based on the difference sequence to obtain a determination result of the stability of the fitting curve; Combine the stability judgment results with the latest fitting curve to output the final water activity measurement value and related data; The specific method for performing denoising, smoothing and formatting on the raw data includes a moving average method or a Kalman filter method; in, The moving average method specifically comprises the following steps: Determine the window size and weight distribution rules based on data characteristics; Arrange the data in ascending order according to the time series by timestamp. Missing values ​​need to be interpolated or removed, and the moving average is calculated after removing obvious outliers. The Kalman filtering method specifically comprises the following steps: Define the initial state vector and its error covariance matrix according to the data type. If the initial state is unknown, it can be set as a diagonal matrix; Predict the state at the next moment based on the system dynamic model and update the uncertainty of the predicted state; The weight coefficients of the fusion prediction and observation are used to correct the predicted state using the actual observation value, and the uncertainty of the corrected state is updated at the same time; The predicted state and state uncertainty are used as input for the next moment, and the prediction and correction process is repeated to achieve real-time recursive filtering; The specific fitting steps in the water activity prediction model are as follows: Initial fitting: After obtaining a small number of discrete data points at the beginning of the measurement, the first fitting curve is generated; Dynamic update: As the test progresses, new data points are continuously added, and the existing water activity prediction model is updated or refitted; Real-time output of prediction results: Based on the latest fitting curve, the water activity prediction value after the current time point is obtained; The specific output steps of the determination result are as follows: Comparison of continuous fitting curves: After obtaining two continuous fitting curves, select several corresponding water activity values ​​y at the same time point x for comparison to obtain a difference sequence; Calculate difference statistics: perform statistical analysis on the difference series, specifically calculate the standard deviation or maximum difference; Threshold determination: When the difference statistic is lower than the preset threshold, the fitting curve is considered to be stable, indicating that the final value of water activity no longer changes significantly; Output results: Once it is determined to be stable, a "measurement stable" signal is sent to the result output module for final result extraction and presentation.

2. A rapid detection method for water activity according to claim 1, characterized in that: The measuring device comprises a dew point method, a resistance method or a microwave absorption method for measuring water activity.

3. A rapid detection method for water activity according to claim 2, characterized in that: The dew point method specifically comprises the following steps: Crush solid samples to a particle size no larger than 5 mm. Liquid or semi-fluid samples need to be stirred evenly, and the sample filling volume should be controlled to two-thirds of the sample cup capacity; The sample cup is quickly placed in the measurement chamber. After sealing, the semiconductor module is used to rapidly reduce the mirror temperature from ambient temperature to below the expected dew point under a constant temperature and humidity environment. The temperature is then raised back up at a rate of 0.1°C / second. The mirror reflectivity is scanned in real time by an infrared laser to capture the moment of condensation. The system automatically records the critical temperature of condensation and calculates the water activity based on the critical temperature and the Clausius-Clapeyron equation.

4. A rapid detection method for water activity according to claim 2, characterized in that, The resistance method specifically comprises the following steps: Crush solid samples to a particle size no larger than 3 mm. Liquid or semi-fluid samples need to be stirred evenly, and the sample filling volume should be controlled to two-thirds of the sample cup capacity; The sample cup is quickly placed into the measuring chamber, sealed, and the sample is absorbed by the hygroscopic material under a constant temperature and humidity environment. The real-time resistance value is measured using a sensor, and the steady-state resistance value is recorded. Based on the steady-state resistance value, the instrument's built-in algorithm is used to convert the steady-state resistance value into a water activity value.

5. A rapid detection method for water activity according to claim 2, characterized in that: The microwave absorption method specifically comprises the following steps: Solid samples should be crushed to a particle size of no more than 3mm. Liquid or semi-fluid samples should be stirred evenly, and the sample filling thickness should be controlled within 2mm-5mm. The sample is placed in a sample dish made of microwave-transmitting material. Microwaves of a fixed frequency are emitted, penetrate the sample, and the attenuation signal is received. The energy loss value of the microwave after passing through the sample is recorded by the sensor. Based on the energy loss value and the correlation between the absorption intensity of microwaves by water and molecular polarity, a mathematical model of loss factor and water activity was established to calculate the water activity.

6. A rapid water activity detection method according to claim 1, characterized in that: The specific steps of the associated storage are as follows: Data reception and marking: Collect and receive water activity measurement values ​​and record the timestamp of the measurement moment. Each data point is composed of a timestamp and a water activity value to identify its time series characteristics. Data storage or caching: The received data points are stored in the system cache or database in chronological order. At the same time, the system transmits the data to the host computer or cloud for backup through the communication interface; Abnormal data filtering: Mark or remove data whose values ​​are outside the reasonable range; or use historical data comparison or statistical methods to identify abnormal points.

7. A rapid water activity detection system, using the rapid water activity detection method according to any one of claims 1 to 6, characterized in that: include: The sensing detection module includes a water activity sensor or a measuring device for sensing and measuring the water activity value of the sample at different time points in real time; a data acquisition module, configured to obtain real-time water activity data from the sensing detection module and associate corresponding timestamps with water activity values ​​for storage; A data processing module, configured to perform denoising, smoothing, and formatting on the raw data collected by the data acquisition module, and to cache the data according to a set time interval or real-time update mode; A data fitting and prediction module is used to perform real-time fitting of the water activity time series curve using a water activity prediction model based on the discrete water activity data output by the data processing module to obtain a fitting curve, and to infer the water activity value at a future moment based on the fitting curve; a stability judgment module, configured to compare the differences between two consecutive fitting curves to obtain a difference sequence, and to perform statistical analysis based on the difference sequence to obtain a judgment result on the stability of the fitting curve; The result output module is used to combine the stability judgment result with the latest fitting curve and output the final water activity measurement value and related data.

Images