Intelligent monitoring method for stem cell liquid nitrogen storage state based on multi-sensor fusion

Abstract

This invention discloses an intelligent monitoring method for the liquid nitrogen storage status of stem cells based on multi-sensor fusion, belonging to the field of stem cell cryogenic storage monitoring technology. The method includes collecting multimodal sensing data streams within the liquid nitrogen storage tank, generating a standard monitoring dataset through spatiotemporal synchronization and outlier removal, and extracting characteristic values ​​of liquid nitrogen level, gas phase pressure, and tank wall temperature field distribution. The liquid level characteristic values ​​are input into a fuzzy logic inference model to output a liquid level deviation index. The pressure and temperature characteristic values ​​are fused to construct a thermodynamic equilibrium equation coupled analysis framework to calculate the liquid nitrogen evaporation rate. The two types of indicators are input into a long short-term memory network model to obtain the estimated empty tank time. Comparison with a preset safety threshold triggers a graded early warning and executes automated liquid replenishment. This method can achieve multi-parameter correlation analysis and time-series prediction, improve the accuracy of storage status determination, optimize anomaly response efficiency, and form an integrated automatic control system of monitoring, early warning, and liquid replenishment, ensuring the stable and controllable state of stem cell liquid nitrogen storage.

Description

Technical Field

[0001] This invention belongs to the field of stem cell cryogenic storage monitoring technology, specifically a method for intelligent monitoring of stem cell liquid nitrogen storage status based on multi-sensor fusion. Background Technology

[0002] Current methods for monitoring the status of stem cell liquid nitrogen storage mainly use a single sensor to collect parameters such as liquid nitrogen level, gas phase pressure, and tank wall temperature. Each type of monitoring parameter is processed independently and compared with a fixed threshold. Liquid level anomalies are directly determined by fixed values. No correlation analysis is conducted between gas phase pressure and tank wall temperature parameters. Liquid nitrogen evaporation status is indirectly estimated based on a single parameter. The remaining safe storage time is estimated by human experience. Anomaly warnings and liquid replenishment operations are all triggered manually on-site.

[0003] Monitoring data collected by a single sensor is susceptible to interference from the cryogenic storage environment, resulting in data deviation. Independent parameter processing mode cannot reflect the inherent correlation characteristics of various state parameters in liquid nitrogen storage. Fixed threshold judgment cannot adapt to the dynamic fluctuations of liquid nitrogen level. Separate analysis of pressure and temperature parameters cannot match the thermodynamic changes in cryogenic liquid nitrogen storage. The remaining safe time estimated manually differs significantly from the actual storage state, which easily leads to problems such as delayed warnings, false alarms, and missed alarms. Manual operation cannot achieve automated connection between monitoring, early warning, and liquid replenishment.

[0004] It is necessary to complete the spatiotemporal synchronization and outlier removal of multimodal sensing data streams, realize the quantitative judgment of liquid level anomalies by relying on fuzzy logic reasoning, construct a pressure-temperature coupled analysis framework to accurately calculate the liquid nitrogen evaporation rate, combine long short-term memory network to complete the prediction of remaining safe time, trigger graded early warning instructions based on the prediction results, and complete the automated liquid replenishment process. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes an intelligent monitoring method for the liquid nitrogen storage status of stem cells based on multi-sensor fusion, including: Acquire multimodal sensing data streams from within liquid nitrogen storage tanks; The multimodal sensing data stream is spatiotemporally synchronized and outlier removed to generate a standard monitoring dataset; Extract the liquid nitrogen level characteristic value, gas phase pressure characteristic value, and tank wall temperature field distribution characteristic value from the standard monitoring dataset; The liquid nitrogen level characteristic value is input into the liquid level anomaly judgment model based on fuzzy logic reasoning, and the liquid level deviation index is output. By integrating the gas phase pressure characteristic value and the tank wall temperature field distribution characteristic value, a pressure-temperature coupling analysis framework based on the thermodynamic equilibrium equation is constructed to calculate the liquid nitrogen evaporation rate. The liquid level deviation index and the liquid nitrogen evaporation rate are input into the remaining safe time prediction model based on a long short-term memory network to generate the estimated empty tank time. The estimated empty tank time is compared with the preset safety threshold to trigger a graded early warning command and execute an automated liquid replenishment process.

[0006] Furthermore, the multimodal sensing data stream is spatiotemporally synchronized and outlier removed to generate a standard monitoring dataset, specifically including: The multimodal sensing data stream includes ultrasonic echo signals, piezoresistive pressure sensor readings, distributed fiber optic temperature sequences, and RFID tag reading records. Based on the timestamp in the RFID tag reading record, linear interpolation alignment is performed on the remaining sensor data; An anomaly detection window based on the Laida criterion is constructed, and the aligned data is traversed to identify and remove impulse noise exceeding three times the standard deviation. Kalman filtering is applied to the data after removing impulse noise to smooth it out and suppress high-frequency jitter. The smoothed data is uniformly mapped to a dimension range of zero to one hundred to form the standard monitoring dataset.

[0007] Further, extracting liquid nitrogen level feature values ​​from the standard monitoring dataset specifically includes: Analyze the ultrasonic echo signal, identify the position of the reflected wave peak on the liquid surface, and calculate the round-trip time difference of the sound wave; The original liquid level height is calculated based on the time difference and the sound velocity temperature correction coefficient. The original liquid level height is compensated for by the curvature of the container cone bottom to generate the compensated liquid level height; The ratio of the compensated liquid level height to the total height of the tank is used as the liquid nitrogen level characteristic value.

[0008] Furthermore, the gas phase pressure characteristic values ​​and tank wall temperature field distribution characteristic values ​​are extracted from the standard monitoring dataset, specifically including: The absolute gas phase pressure value is obtained by reading the piezoresistive pressure sensor and subtracting the influence of local atmospheric pressure. The distributed fiber grating temperature sequence is analyzed to obtain the wavelength drift of multiple temperature measuring points arranged along the axial and radial directions of the tank. The wavelength drift is converted into the corresponding Celsius temperature value to construct a three-dimensional temperature field matrix of the tank wall. The moving average of the absolute gas phase pressure value is calculated as the gas phase pressure characteristic value; The highest temperature value, lowest temperature value, and temperature gradient variance in the three-dimensional temperature field matrix of the tank wall are extracted as the characteristic values ​​of the temperature field distribution of the tank wall.

[0009] Furthermore, the liquid nitrogen level characteristic value is input into a liquid level anomaly determination model based on fuzzy logic reasoning, and the output liquid level deviation index specifically includes: Define a fuzzy linguistic variable for abnormal liquid level, which includes three levels: normal, low, and severely low. Define a membership function for the liquid nitrogen level characteristic value, wherein the membership function is a triangular or trapezoidal distribution; The membership degree of the current liquid nitrogen level characteristic value to each level is calculated based on the membership function. The membership degree is defuzzified using a weighted average method, and a quantized value between zero and one hundred is output as the liquid level deviation index.

[0010] Furthermore, by integrating the gas phase pressure characteristic value and the tank wall temperature field distribution characteristic value, a pressure-temperature coupled analysis framework based on the thermodynamic equilibrium equation is constructed to calculate the liquid nitrogen evaporation rate, specifically including: By using the discretized form of the Clausius-Clapeyron equation, the relationship between saturated vapor pressure and temperature is established. Substituting the characteristic value of the gas phase pressure into the relationship, the theoretical value of the current gas phase temperature can be derived. Calculate the residual between the theoretical value and the highest temperature value among the characteristic values ​​of the tank wall temperature field distribution; The heat transfer coefficient of the evaporation model is dynamically adjusted based on the residual value, and the mass loss of liquid nitrogen per unit time is calculated using the law of conservation of energy. The mass loss is converted into volume loss to obtain the liquid nitrogen evaporation rate.

[0011] Furthermore, the liquid level deviation index and the liquid nitrogen evaporation rate are input into a remaining safety time prediction model based on a long short-term memory network to generate the estimated empty tank time, specifically including: Construct a long short-term memory network structure that includes an input gate, a forget gate, and an output gate; The liquid level deviation index, the liquid nitrogen evaporation rate, and the current liquid level height are used as time-series input sequences and input into the long short-term memory network. The forget gate controls the proportion of the cell state forgotten from the previous moment, and the input gate updates the current cell state. After mapping with a fully connected layer, the predicted curve of the liquid level height for the next 24 hours is output. The predicted empty tank time is obtained by calculating the difference between the horizontal axis of the predicted curve and the preset minimum safe liquid level.

[0012] Furthermore, based on the comparison between the estimated empty tank time and the preset safety threshold, a tiered early warning instruction is triggered, specifically including: Three-level time thresholds are set, including an emergency threshold, an alarm threshold, and a reminder threshold; Determine whether the estimated empty tank time is less than the emergency threshold; If the value is less than the emergency threshold, an emergency command is generated that includes an audible and visual alarm activation signal and a remote communication interruption signal. If the value is not less than the emergency threshold but less than the alarm threshold, an alarm command containing a local buzzer activation signal and an SMS notification signal is generated. If the value is not less than the alarm threshold but less than the reminder threshold, a reminder instruction containing an indicator light flashing signal is generated.

[0013] Furthermore, the automated fluid resuscitation process specifically includes: In response to the emergency command or the alarm command, activate the fluid replenishment pump start relay; Open the solenoid valve of the quick replenishment interface connected to the external liquid nitrogen Dewar; The liquid nitrogen level characteristic value is monitored in real time. When the liquid nitrogen level characteristic value rises back to 90% of the full tank level, a shutdown signal is generated. According to the shutdown signal, the rapid replenishment interface solenoid valve and the replenishment pump start relay are shut off sequentially. Record the volume, duration, and triggering reason of this fluid resuscitation, and write it to the operation log database.

[0014] Furthermore, it also includes steps to ensure the viability of stem cell samples, specifically including: When the temperature gradient variance in the characteristic value of the temperature field distribution of the tank wall exceeds a preset uniformity threshold, it is determined that there is thermal convection disturbance inside the tank. In response to the thermal convection disturbance determination result, the in-tank stirring fan is started or the liquid nitrogen spray angle is adjusted to reduce the temperature gradient variance; Periodically read the RFID tag reading records to verify the timestamps of sample access operations and the identity of the operators; When an unauthorized access operation is detected, the can lid electromagnetic lock is locked and a security audit event is generated. All monitoring data, early warning records, and operation logs are linked and stored with the unique identifier of the sample, forming a data chain that is traceable throughout the entire process.

[0015] Compared with the prior art, the beneficial effects of the present invention are: A fuzzy logic reasoning model is used to construct a liquid level anomaly judgment model. This model processes the liquid nitrogen level characteristic values ​​and outputs a liquid level deviation index, adapting to the dynamic fluctuations of liquid nitrogen levels during storage. It mitigates the judgment errors caused by fixed threshold judgment methods, accurately reflects the degree of deviation between the liquid level state and the normal range, and improves the precision of liquid level anomaly judgment. Furthermore, by integrating gas phase pressure characteristic values ​​and tank wall temperature field distribution characteristic values, a pressure-temperature coupled analysis framework is built based on the thermodynamic balance equation. This enables the correlation calculation of pressure and temperature parameters, closely reflecting the thermodynamic changes of liquid nitrogen within the storage tank. It accurately calculates the liquid nitrogen evaporation rate, avoiding the problem that single-parameter analysis cannot reflect the true changes in the storage state, ensuring that the calculated evaporation rate is consistent with the actual storage state.

[0016] By using liquid level deviation and liquid nitrogen evaporation rate as input data, a remaining safety time prediction model based on a long short-term memory network is constructed. Leveraging the network's temporal feature extraction capabilities, the model can uncover the impact of liquid level deviation and evaporation rate changes over time on the storage status, generating an estimated empty tank time that closely reflects the actual storage process and improving the accuracy of remaining safety time predictions. Comparing the estimated empty tank time with a preset safety threshold can directly trigger tiered early warning commands and drive an automated liquid replenishment process. This achieves seamless execution of monitoring, early warning, and liquid replenishment, reducing response delays caused by manual intervention and creating a complete closed loop for storage status monitoring and anomaly handling. This aligns with the continuous and stable monitoring requirements of stem cell liquid nitrogen storage, making anomaly responses and replenishment operations more consistent with actual operational needs and ensuring the continuity of storage status monitoring and handling. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of the intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion described in this invention. Figure 2 A flowchart for extracting characteristic values ​​of gas phase pressure and tank wall temperature field distribution; Figure 3 The graph shows the membership function curve for the fuzzy logic of liquid nitrogen level. Figure 4 A timing analysis diagram of the automated liquid nitrogen replenishment process for liquid nitrogen storage; Figure 5 This is a time series monitoring graph of the temperature gradient variance of the tank wall. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0019] See Figure 1 This invention provides an intelligent monitoring method for the liquid nitrogen storage status of stem cells based on multi-sensor fusion, the specific method including: The system collects multimodal sensing data streams from the liquid nitrogen storage tank; performs spatiotemporal synchronization and outlier removal on the multimodal sensing data streams to generate a standard monitoring dataset; extracts liquid nitrogen level characteristic values, gas phase pressure characteristic values, and tank wall temperature field distribution characteristic values ​​from the standard monitoring dataset; inputs the liquid nitrogen level characteristic values ​​into a liquid level anomaly judgment model based on fuzzy logic reasoning to output a liquid level deviation index; fuses the gas phase pressure characteristic values ​​and the tank wall temperature field distribution characteristic values ​​to construct a pressure-temperature coupling analysis framework based on the thermodynamic balance equation and calculates the liquid nitrogen evaporation rate; inputs the liquid level deviation index and the liquid nitrogen evaporation rate into a remaining safety time prediction model based on a long short-term memory network to generate the estimated empty tank time; compares the estimated empty tank time with a preset safety threshold to trigger a graded early warning command and execute an automated liquid replenishment process.

[0020] In one embodiment of the present invention, the multimodal sensing data stream includes ultrasonic echo signals, piezoresistive pressure sensor readings, distributed fiber optic temperature sequences, and RFID tag reading records. Using the timestamps in the RFID tag reading records as a reference, the remaining sensor data are linearly interpolated and aligned. An anomaly detection window based on the Laida criterion is constructed, and the aligned data is traversed to identify and remove impulse noise exceeding three times the standard deviation. The data after removing impulse noise is smoothed using Kalman filtering to suppress high-frequency jitter. The smoothed data is then uniformly mapped to a dimension range of zero to one hundred to form a standard monitoring dataset.

[0021] In practice, the multimodal sensing data stream includes ultrasonic echo signals, piezoresistive pressure sensor readings, distributed fiber Bragg grating temperature sequences, and RFID tag reading records. The RFID tag reading records originate from electronic tags attached to the stem cell storage basket. When the basket is accessed, the reader generates a record containing a timestamp accurate to the millisecond level, such as "202X-XX-XX10:05:30.125,Basket_ID_001,Access". In practice, the timestamps in the RFID tag reading records are used as a benchmark to perform linear interpolation alignment on the data from other sensors. The ultrasonic echo signal, the piezoresistive pressure sensor readings, and the distributed fiber grating temperature sequence each have independent clock sources that may have slight drift. By using the timestamps in the RFID tag reading records as a common reference point, the data sequences from other sensors are resampled onto this unified time axis. Assuming that the original liquid level value collected by the ultrasonic sensor at 10:05:30.100 is 850 mm, and the value collected at 10:05:30.150 is 849.8 mm, while the target alignment timestamp is 10:05:30.125, the aligned liquid level value is calculated as 849.9 mm by performing linear interpolation on the two points. Every data point in the piezoresistive pressure sensor readings and the distributed fiber grating temperature sequence is processed according to this principle to ensure that all data are strictly synchronized in the time dimension.

[0022] In some embodiments, an anomaly detection window based on the Laida criterion is constructed, and the aligned data is traversed to identify and remove impulse noise exceeding three times the standard deviation. For a data window containing N consecutive sampling points, the arithmetic mean μ and standard deviation σ of the data within the window are calculated. Data points that meet the following conditions are identified and removed: in: This represents the value of the i-th data point within the window. This represents the arithmetic mean of the data window. This represents the standard deviation of the data window. For example, in a short sequence of liquid level data of length 100, with a mean μ of 500.0 mm and a standard deviation σ of 0.5 mm, any data point with a value less than 498.5 mm or greater than 501.5 mm will be considered impulse noise and removed. The positions of the removed points are filled using the linear interpolation results of the preceding and following valid data. In some embodiments, the data after removing impulse noise is smoothed by Kalman filtering to suppress high-frequency jitter. Kalman filtering is performed iteratively through two steps: state prediction and measurement update. Its state vector can contain the liquid level value and its rate of change. By setting the process noise covariance matrix and the measurement noise covariance matrix, the filter can effectively smooth random fluctuations in the data. For example, a small high-frequency jitter in the liquid level caused by slight boiling in the tank will be suppressed after Kalman filtering, outputting a smooth trend curve.

[0023] Optionally, the smoothed data can be uniformly mapped to a dimension range of 0 to 100, forming a standard monitoring dataset. The mapping process uses a linear scaling method. For each type of sensor data, its effective measurement range in the actual physical world needs to be pre-defined. For example, the range of an ultrasonic level gauge is 0 to 1500 mm, the range of a piezoresistive pressure sensor is 0 to 200 kPa, and the range of a distributed fiber Bragg grating temperature sensor is -200 to 50 degrees Celsius. For any smoothed raw data value... Its corresponding standard monitoring data value Calculated using the following formula: in: and These represent the sensor's preset minimum and maximum ranges, respectively. Through this step, sensor data with different physical meanings and units are converted into uniform, dimensionless values ​​between zero and one hundred. For example, a liquid level value of 750 mm is converted to 50.0, a pressure value of 100 kPa is converted to 50.0, and a temperature value of -75 degrees Celsius is also converted to 50.0. This normalization process provides a dimensionally consistent basis for subsequent multi-feature fusion analysis. It can be understood that the entire preprocessing workflow sequentially completes time synchronization, gross error removal, random noise smoothing, and dimensional normalization. The final output standard monitoring dataset is a multi-channel time-series data set with strictly aligned timestamps, smoothed values, and values ​​within the same range.

[0024] In one embodiment of the present invention, ultrasonic echo signals from a standard monitoring dataset are analyzed to identify the position of the reflected wave peaks on the liquid surface and to calculate the round-trip time difference of the sound waves. Based on the time difference and a sound velocity-temperature correction coefficient, the original liquid level height is calculated. The original liquid level height is then compensated for by the container's conical bottom curvature to generate a compensated liquid level height. The ratio of the compensated liquid level height to the total height of the tank is used as the liquid nitrogen level characteristic value. (See also...) Figure 2 The absolute gas phase pressure value is obtained by reading the piezoresistive pressure sensor readings from the standard monitoring dataset and subtracting the influence of local atmospheric pressure. The distributed fiber optic grating temperature sequence in the standard monitoring dataset is analyzed to obtain the wavelength drift of multiple temperature measurement points arranged along the axial and radial directions of the tank. The wavelength drift is converted into corresponding Celsius temperature values ​​to construct a three-dimensional temperature field matrix for the tank wall. The moving average of the absolute gas phase pressure values ​​is calculated as the gas phase pressure characteristic value. The highest temperature value, lowest temperature value, and temperature gradient variance are extracted from the three-dimensional temperature field matrix of the tank wall as the temperature field distribution characteristic values ​​of the tank wall.

[0025] In practice, liquid nitrogen level characteristic values, gas phase pressure characteristic values, and tank wall temperature field distribution characteristic values ​​are extracted from the standard monitoring dataset. The ultrasonic echo signal in the standard monitoring dataset is analyzed to identify the position of the reflection peak on the liquid surface and calculate the round-trip time difference of the sound wave. The ultrasonic echo signal in the standard monitoring dataset is a series of digitized voltage amplitude sequences, corresponding to the time history of ultrasonic wave transmission and reception. By finding the pulse peak with the largest amplitude in the echo signal that exceeds the set threshold, the difference between the corresponding time point and the ultrasonic wave transmission time point is the round-trip time difference of the sound wave. For example, in a measurement, the transmitted pulse is recorded at time point 0 microseconds, and the identified reflection peak on the liquid surface is located at time point 23500 microseconds, then the round-trip time difference of the sound wave is 23.5 milliseconds. In practice, the original liquid level height is calculated based on the time difference and the sound velocity temperature correction coefficient. The propagation speed of ultrasound in the gas phase space is not constant, but varies with the gas temperature. The sound velocity temperature correction coefficient describes the relationship between sound velocity and temperature. The gas phase space temperature of the liquid nitrogen storage tank can be obtained through the upper temperature measurement point in the distributed fiber optic grating temperature sequence. Combining the time difference and the corrected sound velocity, the original liquid level height from the probe to the liquid surface is calculated using the formula "distance = sound velocity × time / 2". For example, when the gas phase temperature is -150 degrees Celsius, the corresponding sound velocity is found to be 220 m / s by referring to the table. Combining the time difference of 23.5 milliseconds, the calculated original liquid level height is 2.585 meters. In practice, the original liquid level is compensated for by the conical bottom curvature of the container to generate a compensated liquid level. The bottom of a liquid nitrogen storage tank is typically conical or spherical. The original liquid level obtained by ultrasonic ranging is the vertical distance from the probe to the liquid surface, not the actual volume of liquid inside the tank. Compensation calculations need to be performed based on the geometric dimensions of the tank bottom. For example, for a tank with a hemispherical bottom, the compensation amount is a function of the liquid level height and the bottom radius. The compensated liquid level is the original liquid level minus the compensation value calculated based on the geometric model. In practice, the ratio of the compensated liquid level to the total tank height is used as the liquid nitrogen level characteristic value. The total tank height is a known fixed design parameter, such as 3.0 meters. If the compensated liquid level is 2.4 meters, the calculated liquid nitrogen level characteristic value will be 0.8 or 80%.

[0026] In some embodiments, the piezoresistive pressure sensor readings in the standard monitoring dataset are read, and the influence of local atmospheric pressure is subtracted to obtain the absolute gas phase pressure value. The piezoresistive pressure sensor reading is usually a gauge pressure value, i.e., the difference between the pressure inside the tank and the local atmospheric pressure. The absolute gas phase pressure value is equal to the gauge pressure reading plus the local atmospheric pressure value measured or set in real time by a barometer. For example, if the piezoresistive pressure sensor reading is 15 kPa and the local atmospheric pressure is 101 kPa, then the calculated absolute gas phase pressure value is 116 kPa. In some embodiments, the distributed fiber grating temperature sequence in the standard monitoring dataset is analyzed to obtain the wavelength drift of multiple temperature measurement points arranged along the axial and radial directions of the tank. The distributed fiber grating temperature sequence is a set of values ​​in wavelength units, reflecting the offset of the center wavelength of each fiber grating sensor. The wavelength drift is the difference between the currently measured center wavelength value and the calibrated wavelength of the sensor at zero stress and reference temperature. Optionally, the wavelength drift is converted into the corresponding Celsius temperature value to construct a three-dimensional temperature field matrix of the tank wall. The temperature value of each fiber grating temperature measurement point is calculated using the following formula: in: This represents the calculated Celsius temperature value. This represents the wavelength shift obtained from the temperature sequence of the distributed fiber Bragg grating. The temperature sensitivity coefficient represents the fiber Bragg grating sensor, a constant determined during sensor calibration. The temperature values ​​calculated from all temperature measurement points at different axial heights and radial azimuths of the tank are then filled into a three-dimensional array according to their preset spatial coordinates, forming the three-dimensional temperature field matrix of the tank wall. This matrix visually reflects the temperature distribution on the tank wall surface. Optionally, the moving average of the absolute gas phase pressure values ​​can be calculated as the gas phase pressure characteristic value. For example, the arithmetic mean of the absolute gas phase pressure values ​​from the most recent 100 sampling times can be used as the current gas phase pressure characteristic value, which can smooth out short-term fluctuations in the pressure signal. It is understandable that the highest temperature value, lowest temperature value, and temperature gradient variance in the three-dimensional temperature field matrix of the tank wall are extracted as characteristic values ​​of the tank wall temperature field distribution. The highest and lowest temperature values ​​are directly selected from all elements of the three-dimensional temperature field matrix, while the temperature gradient variance needs to be calculated. The process involves first calculating the temperature difference between adjacent points in the axial and radial directions of the three-dimensional temperature field matrix to form a temperature gradient matrix, and then calculating the variance of all elements in this gradient matrix. The temperature gradient variance reflects the uniformity of the tank wall temperature distribution. It is also understood that through the above steps, liquid nitrogen level characteristic values, gas phase pressure characteristic values, and tank wall temperature field distribution characteristic values ​​including the highest temperature, lowest temperature, and temperature gradient variance are extracted from a unified standard monitoring dataset.

[0027] In one embodiment of the invention, a fuzzy linguistic variable for abnormal liquid level is defined, comprising three levels: normal, low, and severely low. A membership function for the liquid nitrogen level characteristic value is set, with the membership function being a triangular or trapezoidal distribution. The membership degree of the current liquid nitrogen level characteristic value to each level is calculated based on the membership function. A weighted average method is used to defuzzify the membership degree, outputting a quantized value between zero and one hundred as the liquid level deviation index. The discretized form of the Clausius-Clapeyron equation is used to establish the relationship between saturated vapor pressure and temperature. The vapor phase pressure characteristic value is substituted into this relationship to deduce the theoretical value of the current vapor phase temperature. The residual between the theoretical value and the highest temperature value in the tank wall temperature field distribution characteristic value is calculated. The heat transfer coefficient of the evaporation model is dynamically adjusted based on the magnitude of the residual, and the mass loss of liquid nitrogen per unit time is calculated using the law of conservation of energy. The mass loss is converted into volume loss to obtain the liquid nitrogen evaporation rate.

[0028] In practical implementation, the liquid level anomaly judgment model and the calculation of liquid nitrogen evaporation rate define a fuzzy linguistic variable for liquid level anomalies, which includes three levels: normal, low, and severely low. A membership function for the liquid nitrogen level characteristic value is set, with a triangular or trapezoidal distribution. The liquid nitrogen level characteristic value is a ratio based on the normalized total height of the tank, and its universe of discourse is set to 0 to 1. For the "normal" level, a triangular membership function can be used, with its core parameters including the vertices and the boundary points of the support interval. Refer to Table 1, which shows an example of a possible membership function parameter setting.

[0029] Table 1: Membership Function Parameter Settings Table In practice, the membership degree of the current liquid nitrogen level characteristic value to each level is calculated based on the membership function. For a specific liquid nitrogen level characteristic value, such as 0.65, the membership degree of its belonging to the "normal" level is calculated according to the parameters in the table above. This involves determining whether 0.65 falls within the trapezoidal core interval [0.60, 0.80] and the right hypotenuse interval [0.80, 1.00]. The trapezoidal membership function calculation formula is applied, and the membership degree of its belonging to the "low" and "severely low" levels is also calculated. The membership degree is defuzzified using the weighted average method, and a quantitative value between zero and one hundred is output as the liquid level deviation index. A representative value is preset for each fuzzy linguistic variable level, for example, 0 for "normal", 50 for "low", and 100 for "severely low". The calculated membership degree of each level is multiplied by the corresponding representative value, summed, and then divided by the sum of the membership degrees to obtain the final liquid level deviation index.

[0030] In some embodiments, the characteristic values ​​of gas phase pressure and the characteristic values ​​of tank wall temperature field distribution are integrated to construct a pressure-temperature coupled analysis framework based on the thermodynamic equilibrium equation to calculate the liquid nitrogen evaporation rate. The discretized form of the Clausius-Clapeyron equation is used to establish the relationship between saturated vapor pressure and temperature. Substituting the characteristic values ​​of gas phase pressure into the relationship, the theoretical value of the current gas phase temperature is derived. It can be understood that the residual is calculated between the theoretical value and the highest temperature value in the tank wall temperature field distribution characteristic values. The theoretical value is the saturation temperature calculated based on pressure, while the highest temperature value is the measured maximum value in the three-dimensional temperature field matrix of the tank wall; the difference between the two is the residual. It can be understood that the heat transfer coefficient of the evaporation model is dynamically adjusted according to the magnitude of the residual. The residual reflects the degree of deviation between the actual heat exchange state and the ideal saturation state. The adjustment of the heat transfer coefficient K follows a preset mapping relationship. For example, when the absolute value of the residual is less than 1 degree Celsius, K takes the base value; when the absolute value of the residual is between 1 and 5 degrees Celsius, the value of K increases linearly. The mass loss of liquid nitrogen per unit time can be calculated using the law of conservation of energy. The heat absorbed during evaporation is equal to the heat transferred through the tank wall. The heat transfer formula is: in: Represents the amount of heat transferred per unit time. This represents the dynamically adjusted heat transfer coefficient. Represents the effective heat transfer area. This represents the highest temperature value among the characteristic values ​​of the temperature field distribution on the tank wall. This represents the theoretical value of the gas phase temperature derived from the characteristic value of the gas phase pressure. It also represents the amount of heat transferred. Divided by the latent heat of vaporization of liquid nitrogen That is, to obtain the mass loss rate per unit time. By converting mass loss into volume loss, the evaporation rate of liquid nitrogen and the density of liquid nitrogen can be obtained. It is known that the volumetric evaporation rate can be obtained by dividing the mass loss rate by the density. .

[0031] See Figure 3This is a fuzzy logic membership function curve for liquid nitrogen level. It intuitively presents the core mechanism of fuzzy logic reasoning: mapping continuous liquid level values ​​to the membership degrees of three fuzzy linguistic variables—"normal / low / severely low"—solving the fuzziness problem in liquid level safety determination. It clearly demonstrates the design logic of the trapezoidal / triangular membership function in the patent, verifying the rationality of the parameter settings. Through the visualization of example liquid levels, the calculation correctness of the fuzzy logic model can be quickly verified, facilitating algorithm debugging and parameter optimization. On-site maintenance personnel can quickly understand the liquid level determination logic through this diagram without delving into mathematical formulas, lowering the operational threshold. As a visual support for the core algorithm, it clearly showcases the invention's points, enhancing the technical transparency of the patent application. It provides an intuitive basis for subsequent calculations of liquid level deviation indicators and is a core prerequisite for predicting remaining safe time and issuing tiered early warnings.

[0032] In one embodiment of the present invention, a long short-term memory (LSTM) network structure including an input gate, a forget gate, and an output gate is constructed. The liquid level deviation index, liquid nitrogen evaporation rate, and current liquid level height are used as temporal input sequences and input into the LSM network. The forget gate controls the forgetting ratio of the cell state at the previous moment, and the input gate updates the current cell state. After mapping through a fully connected layer, a predicted curve of the liquid level height for the next 24 hours is output. The difference between the predicted curve and the preset minimum safe liquid level is calculated to obtain the estimated empty tank time. Three time thresholds are set, including an emergency threshold, an alarm threshold, and a reminder threshold. It is determined whether the estimated empty tank time is less than the emergency threshold. If it is less than the emergency threshold, an emergency command including an audible and visual alarm activation signal and a remote communication interruption signal is generated. If it is not less than the emergency threshold but less than the alarm threshold, an alarm command including a local buzzer activation signal and an SMS notification signal is generated. If it is not less than the alarm threshold but less than the reminder threshold, a reminder command including an indicator light flashing signal is generated. In response to the emergency command or alarm command, the replenishment pump start relay is activated. Open the solenoid valve of the quick replenishment interface connected to the external liquid nitrogen Dewar. Monitor the liquid nitrogen level characteristic value in real time. When the liquid nitrogen level characteristic value rises back to 90% of the full tank level, generate a shutdown signal. Based on the shutdown signal, sequentially close the quick replenishment interface solenoid valve and the replenishment pump start relay. Record the volume, duration, and triggering reason of this replenishment, and write it to the operation log database.

[0033] In the specific implementation, the remaining safety time prediction, graded early warning, and automated liquid replenishment are achieved by constructing a Long Short-Term Memory (LSTM) network structure containing an input gate, a forget gate, and an output gate. The number of hidden layer units in the LTM network is set to 128, and the network contains three stacked LTM layers. In the specific implementation, the liquid level deviation index, liquid nitrogen evaporation rate, and current liquid level height are used as temporal input sequences into the LTM network. The temporal input sequence is a two-dimensional matrix arranged in chronological order, where each row represents a sampling time and each column represents a feature. The three features are arranged in the order of liquid level deviation index, liquid nitrogen evaporation rate, and current liquid level height. The sequence length is set to the data from the most recent 120 consecutive sampling times. The forget gate controls the proportion of cell states forgotten from the previous time step, and the input gate updates the current cell state. The calculation of the forget gate and the input gate involves the sigmoid activation function, used to generate a gating value between 0 and 1. The cell state update follows the classic calculation process of the LTM network, i.e., the forget gate determines how much of the cell state from the previous time step is retained, and the input gate determines how many new candidate cell states are added. After mapping by a fully connected layer, a predicted curve for the liquid level height over the next 24 hours is output. The output of the last hidden layer of the Long Short-Term Memory network is fed into a fully connected layer with 24 neurons, each neuron outputting a predicted liquid level height value for the next 24 hours. Connecting these points forms the predicted curve. The difference between the predicted curve and the preset minimum safe liquid level is calculated to obtain the estimated empty tank time. The preset minimum safe liquid level is a constant, such as 20% of the total tank height. On the generated predicted liquid level height curve for the next 24 hours, the time point corresponding to the first time the curve falls below or equals the minimum safe liquid level value is found. The time difference between this time point and the current time is the estimated empty tank time. For example, if the current time is 0 hours and the predicted curve crosses the minimum safe liquid level at 18.5 hours, then the estimated empty tank time is 18.5 hours.

[0034] In some embodiments, a tiered early warning command is triggered by comparing the estimated empty tank time with a preset safety threshold. Three time thresholds are set, including an emergency threshold, an alarm threshold, and a reminder threshold. The specific values ​​of these three thresholds can be set according to the storage security policy; for example, the reminder threshold could be set to 48 hours, the alarm threshold to 24 hours, and the emergency threshold to 4 hours. It is then determined whether the estimated empty tank time is less than the emergency threshold. If the estimated empty tank time is less than the emergency threshold, an emergency command is generated, including an audible and visual alarm activation signal and a remote communication interruption signal. The audible and visual alarm activation signal activates a high-decibel alarm bell and rotating warning light on-site. The remote communication interruption signal triggers the sending of an alarm SMS message containing an "emergency" identifier to a preset administrator's mobile phone and automatically dials an alarm phone number. If the empty tank time is not less than the emergency threshold but less than the alarm threshold, an alarm command is generated, including a local buzzer activation signal and an SMS notification signal. The local buzzer activation signal causes the device's local buzzer to sound intermittently, and the SMS notification signal triggers the sending of a reminder SMS message containing an "alarm" identifier to a preset administrator's mobile phone. If the value is not less than the alarm threshold but less than the alert threshold, an alert command is generated that includes a flashing indicator light signal. The flashing indicator light signal drives the yellow warning indicator light on the device panel to flash at a fixed frequency. Refer to Table 2 for an example of an alert grading rule.

[0035] Table 2: Example Table of Early Warning Classification Rules It is understood that, in executing the automated replenishment process, in response to emergency or alarm commands, the replenishment pump start relay is activated. The control circuit outputs a high-level signal to the coil of the replenishment pump start relay, causing the relay contacts to close and thus connecting the main power circuit of the replenishment pump. The solenoid valve of the quick replenishment interface connected to the external liquid nitrogen Dewar is opened. The control circuit sends an opening signal to the output terminal of the quick replenishment interface solenoid valve, energizing the solenoid valve coil and opening the valve, allowing liquid nitrogen to flow from the external Dewar into the storage tank through the pipeline. The liquid nitrogen level characteristic value is monitored in real time. When the liquid nitrogen level characteristic value rises to 90% of the full tank level, a shutdown signal is generated. The liquid nitrogen level characteristic value is continuously calculated during the replenishment process, and when its value reaches or exceeds a preset threshold of 0.9 (i.e., 90%), the shutdown signal is triggered. Optionally, the quick replenishment interface solenoid valve and the replenishment pump start relay are closed sequentially according to the shutdown signal. The control logic first sends a shutdown signal to the quick replenishment interface solenoid valve, delays for several seconds, and then disconnects the control signal to the replenishment pump start relay to ensure a smooth release of pressure within the pipeline. The volume, duration, and triggering reason of this replenishment are recorded and written to the operation log database. The replenishment volume is obtained by integrating the liquid level change and tank volume curve before and after replenishment is started. The duration is the time interval from activating the replenishment pump start relay to deactivating the replenishment pump start relay. The triggering reason is recorded as "alarm" or "emergency". This information, along with the timestamp, is written to the structured operation log database table.

[0036] See Figure 4 This is a time-series analysis diagram of the automated liquid nitrogen replenishment process for stem cell liquid nitrogen storage tanks, showcasing the entire process of automated replenishment based on multi-sensor fusion. It fully presents the liquid level changes and system status logic throughout the entire cycle, from triggering an anomaly to completing replenishment. From 0-20 minutes, the liquid level is approximately 60%, in a "standby" state, with normal system monitoring. It slowly decreases over time, reaching approximately 35% at 20 minutes, below the minimum safe level (20%), triggering the system's anomaly detection. At 20 minutes, the system responds to the anomaly command, activating the replenishment pump and solenoid valve, entering the "replenishing" state. The liquid level rapidly recovers, jumping to approximately 90% at 40 minutes, reaching the replenishment stop threshold. At 40 minutes, the system generates a shutdown signal, sequentially shutting down the solenoid valve and replenishment pump, switching to the "complete" state. The liquid level remains stable at approximately 85%, the replenishment process ends, and the tank returns to a safe monitoring state.

[0037] In one embodiment of the present invention, when the temperature gradient variance in the characteristic value of the temperature field distribution on the tank wall exceeds a preset uniformity threshold, it is determined that there is a thermal convection disturbance inside the tank. In response to the thermal convection disturbance determination result, the tank's stirring fan is activated or the liquid nitrogen spray angle is adjusted to reduce the temperature gradient variance. The RFID tag reading records are periodically read to verify the timestamps of sample access operations and the operator's identity. When an unauthorized access operation is detected, the tank lid electromagnetic lock is locked and a security audit event is generated. All monitoring data, early warning records, and operation logs are associated and stored with the sample's unique identifier, forming a fully traceable data chain.

[0038] In practical implementation, the stem cell sample viability assurance step determines the presence of thermal convection disturbance within the tank when the temperature gradient variance in the characteristic value of the temperature field distribution on the tank wall exceeds a preset uniformity threshold. The temperature gradient variance is calculated based on the three-dimensional temperature field matrix of the tank wall. The uniformity threshold is a pre-set value, such as 0.5 square degrees Celsius. The current temperature gradient variance is calculated and compared with this threshold. If the calculated variance is 0.8 square degrees Celsius, exceeding the 0.5 square degree Celsius threshold, the system determines that there is significant thermal convection disturbance within the tank. In response to the thermal convection disturbance determination, the system activates the in-tank stirring fan or adjusts the liquid nitrogen spray angle to reduce the temperature gradient variance. After the determination is confirmed, the control unit sends a start command to the drive circuit of the miniature low-speed stirring fan installed inside the tank, or sends an adjustment pulse to the stepper motor controlling the liquid nitrogen spray angle. The goal is to promote uniform mixing of the gas or liquid phases within the tank and reduce local temperature differences.

[0039] In some embodiments, the system periodically reads RFID tag reading records to verify the timestamp of sample access operations and the operator's identity. RFID tag reading records are stored in the system's cache. The system reads and parses the latest records at fixed intervals, such as every minute. Each record contains a unique identifier for the sample basket, an operation timestamp, and the operator's ID card obtained through near-field communication. During verification, the operation timestamp is compared with the system time to confirm the real-time nature of the operation, and the operator's ID card is matched against a pre-stored list of authorized personnel in the database. When an unauthorized access operation is detected, the can lid's electromagnetic lock is locked and a security audit event is generated. Unauthorized access operations include, but are not limited to: the ID card ID not being in the authorized list, or the operation occurring within a preset prohibited time period. For example, if an access operation is detected with the operator's ID card ID "EMP_005", but the authorized list only contains "EMP_001" to "EMP_004", it is determined to be an unauthorized operation. The system immediately sends a locking signal to the can lid's electromagnetic lock to keep it closed, and simultaneously generates an event record in the security audit log containing the event time, illegal ID, and associated sample identifier.

[0040] It is understandable that all monitoring data, early warning records, and operation logs are linked and stored with the unique sample identifier, forming a fully traceable data chain. Monitoring data includes time series such as liquid nitrogen level characteristics, gas phase pressure characteristics, and tank wall temperature field distribution characteristics. Early warning records include the trigger time and type of early warning commands at all levels. Operation logs include records of each liquid replenishment operation and all sample access records. In practice, each operation or status read of the storage tank attempts to associate it with the currently active sample identifier. The active status is usually determined by the most recent RFID tag read record of the basket containing the sample. Optionally, this association can be achieved by establishing a common "sample batch identifier" field and a "timestamp" field in the database table, logically linking data entries from different sources. For example, if the timestamp of a fluid resuscitation operation record is "202X-XX-XX 14:30:00", the system will retrieve the sample identifier codes that were active before and after this time, such as "Batch_2024_ABC", and write this identifier code into the associated field of the fluid resuscitation operation log, thereby establishing a binding relationship between environmental monitoring data, equipment operation logs, and specific stem cell sample batches. In some embodiments, the specific formula for calculating the temperature gradient variance is as follows: in: Represents the variance of the temperature gradient. This represents the total number of temperature gradient vectors calculated from the three-dimensional temperature field matrix of the tank wall. Representing the The magnitude of a temperature gradient vector, Representing all The arithmetic mean of the magnitudes of the temperature gradient vectors. The temperature difference between adjacent temperature measurement points along the axial and radial directions of the three-dimensional temperature field matrix of the tank wall constitutes the gradient vector, and its magnitude characterizes the degree of drastic change in local temperature. The variance is obtained by calculating the dispersion of all gradient magnitudes relative to their mean. .

[0041] See Figure 5This is a time-series monitoring chart of the temperature gradient variance within the tank. Over the 0-24 hour period, the temperature gradient variance consistently remained in the high range of 3.4-4.5, far exceeding the uniformity threshold of 0.5, indicating a persistent and significant thermal convection disturbance within the tank. The variance exhibited periodic fluctuations, with peaks occurring at the 3rd hour (approximately 4.5), 7th hour (approximately 4.4), and 9th hour (approximately 4.35), reflecting the dynamic changes in heat exchange within the tank. Under this operating condition, the system continuously triggers thermal convection disturbance alarms, activating the stirring fan or adjusting the liquid nitrogen spray angle to reduce the temperature gradient variance and ensure the temperature uniformity of the stem cell storage environment. Temperature gradient variance is a core quantitative indicator of the temperature field uniformity in liquid nitrogen storage tanks, directly affecting the storage viability of stem cells. Uneven temperature distribution can lead to localized temperature increases, accelerating liquid nitrogen evaporation and even causing sample inactivation.

[0042] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent monitoring of stem cell liquid nitrogen storage status based on multi-sensor fusion, characterized in that, include: Acquire multimodal sensing data streams from within liquid nitrogen storage tanks; The multimodal sensing data stream is spatiotemporally synchronized and outlier removed to generate a standard monitoring dataset; Extract the liquid nitrogen level characteristic value, gas phase pressure characteristic value, and tank wall temperature field distribution characteristic value from the standard monitoring dataset; The liquid nitrogen level characteristic value is input into the liquid level anomaly judgment model based on fuzzy logic reasoning, and the liquid level deviation index is output. By integrating the gas phase pressure characteristic value and the tank wall temperature field distribution characteristic value, a pressure-temperature coupling analysis framework based on the thermodynamic equilibrium equation is constructed to calculate the liquid nitrogen evaporation rate. The liquid level deviation index and the liquid nitrogen evaporation rate are input into the remaining safe time prediction model based on a long short-term memory network to generate the estimated empty tank time. The estimated empty tank time is compared with the preset safety threshold to trigger a graded early warning command and execute an automated liquid replenishment process.

2. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 1, characterized in that, The multimodal sensing data stream is spatiotemporally synchronized and outlier removed to generate a standard monitoring dataset, specifically including: The multimodal sensing data stream includes ultrasonic echo signals, piezoresistive pressure sensor readings, distributed fiber optic temperature sequences, and RFID tag reading records. Based on the timestamp in the RFID tag reading record, linear interpolation alignment is performed on the remaining sensor data; An anomaly detection window based on the Laida criterion is constructed, and the aligned data is traversed to identify and remove impulse noise exceeding three times the standard deviation. Kalman filtering is applied to the data after removing impulse noise to smooth it out and suppress high-frequency jitter. The smoothed data is uniformly mapped to a dimension range of zero to one hundred to form the standard monitoring dataset.

3. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 2, characterized in that, Extracting liquid nitrogen level feature values ​​from the standard monitoring dataset specifically includes: Analyze the ultrasonic echo signal, identify the position of the reflected wave peak on the liquid surface, and calculate the round-trip time difference of the sound wave; The original liquid level height is calculated based on the time difference and the sound velocity temperature correction coefficient. The original liquid level height is compensated for by the curvature of the container cone bottom to generate the compensated liquid level height; The ratio of the compensated liquid level height to the total height of the tank is used as the liquid nitrogen level characteristic value.

4. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 3, characterized in that, Extracting the gas phase pressure characteristic values ​​and tank wall temperature field distribution characteristic values ​​from the standard monitoring dataset, specifically including: The absolute gas phase pressure value is obtained by reading the piezoresistive pressure sensor and subtracting the influence of local atmospheric pressure. The distributed fiber grating temperature sequence is analyzed to obtain the wavelength drift of multiple temperature measuring points arranged along the axial and radial directions of the tank. The wavelength drift is converted into the corresponding Celsius temperature value to construct a three-dimensional temperature field matrix of the tank wall. The moving average of the absolute gas phase pressure value is calculated as the gas phase pressure characteristic value; The highest temperature value, lowest temperature value, and temperature gradient variance in the three-dimensional temperature field matrix of the tank wall are extracted as the characteristic values ​​of the temperature field distribution of the tank wall.

5. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 4, characterized in that, The liquid nitrogen level characteristic value is input into the liquid level anomaly determination model based on fuzzy logic reasoning, and the output liquid level deviation index includes: Define a fuzzy linguistic variable for abnormal liquid level, which includes three levels: normal, low, and severely low. Define a membership function for the liquid nitrogen level characteristic value, wherein the membership function is a triangular or trapezoidal distribution; The membership degree of the current liquid nitrogen level characteristic value to each level is calculated based on the membership function. The membership degree is defuzzified using a weighted average method, and a quantized value between zero and one hundred is output as the liquid level deviation index.

6. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 5, characterized in that, By integrating the gas phase pressure characteristic values ​​and the tank wall temperature field distribution characteristic values, a pressure-temperature coupled analysis framework based on the thermodynamic equilibrium equation is constructed to calculate the liquid nitrogen evaporation rate, specifically including: By using the discretized form of the Clausius-Clapeyron equation, the relationship between saturated vapor pressure and temperature is established. Substituting the characteristic value of the gas phase pressure into the relationship, the theoretical value of the current gas phase temperature can be derived. Calculate the residual between the theoretical value and the highest temperature value among the characteristic values ​​of the tank wall temperature field distribution; The heat transfer coefficient of the evaporation model is dynamically adjusted based on the residual value, and the mass loss of liquid nitrogen per unit time is calculated using the law of conservation of energy. The mass loss is converted into volume loss to obtain the liquid nitrogen evaporation rate.

7. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 6, characterized in that, The liquid level deviation index and the liquid nitrogen evaporation rate are input into a remaining safety time prediction model based on a long short-term memory network to generate the estimated empty tank time, specifically including: Construct a long short-term memory network structure that includes an input gate, a forget gate, and an output gate; The liquid level deviation index, the liquid nitrogen evaporation rate, and the current liquid level height are used as time-series input sequences and input into the long short-term memory network. The forget gate controls the proportion of the cell state forgotten from the previous moment, and the input gate updates the current cell state. After mapping with a fully connected layer, the predicted curve of the liquid level height for the next 24 hours is output. The predicted empty tank time is obtained by calculating the difference between the horizontal axis of the predicted curve and the preset minimum safe liquid level.

8. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 7, characterized in that, Based on the comparison between the estimated empty tank time and the preset safety threshold, a tiered early warning instruction is triggered, specifically including: Three-level time thresholds are set, including an emergency threshold, an alarm threshold, and a reminder threshold; Determine whether the estimated empty tank time is less than the emergency threshold; If the value is less than the emergency threshold, an emergency command is generated that includes an audible and visual alarm activation signal and a remote communication interruption signal. If the value is not less than the emergency threshold but less than the alarm threshold, an alarm command containing a local buzzer activation signal and an SMS notification signal is generated. If the value is not less than the alarm threshold but less than the reminder threshold, a reminder instruction containing an indicator light flashing signal is generated.

9. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 8, characterized in that, The automated fluid resuscitation process specifically includes: In response to the emergency command or the alarm command, activate the fluid replenishment pump start relay; Open the solenoid valve of the quick replenishment interface connected to the external liquid nitrogen Dewar; The liquid nitrogen level characteristic value is monitored in real time. When the liquid nitrogen level characteristic value rises back to 90% of the full tank level, a shutdown signal is generated. According to the shutdown signal, the rapid replenishment interface solenoid valve and the replenishment pump start relay are shut off sequentially. Record the volume, duration, and triggering reason of this fluid resuscitation, and write it to the operation log database.

10. The intelligent monitoring method for stem cell liquid nitrogen storage status based on multi-sensor fusion as described in claim 9, characterized in that, It also includes steps to ensure the viability of stem cell samples, specifically including: When the temperature gradient variance in the characteristic value of the temperature field distribution of the tank wall exceeds a preset uniformity threshold, it is determined that there is thermal convection disturbance inside the tank. In response to the thermal convection disturbance determination result, the in-tank stirring fan is started or the liquid nitrogen spray angle is adjusted to reduce the temperature gradient variance; Periodically read the RFID tag reading records to verify the timestamps of sample access operations and the identity of the operators; When an unauthorized access operation is detected, the can lid electromagnetic lock is locked and a security audit event is generated. All monitoring data, early warning records, and operation logs are linked and stored with the unique identifier of the sample, forming a data chain that is traceable throughout the entire process.

Images