A comprehensive performance testing system for medical ultra-low temperature refrigerators

Abstract

The present application relates to the technical field of on-line detection of medical ultra-low temperature freezer performance, and discloses a medical ultra-low temperature freezer performance comprehensive detection system, which comprises a sensing and collecting module outputting synchronous time sequence data flow; a data preprocessing module, which uses the Grubbs criterion to eliminate temperature abnormal values and interpolates to complete, and generates a pretreatment data set by low-pass filtering the compressor electrical parameters; a state classification module, which divides three types of operation time periods, i.e. steady-state insulation, transient cooling and door opening disturbance recovery, in combination with the door switch level and temperature change rate; a parameter identification module, which solves the equivalent thermal resistance, equivalent heat capacity and refrigeration efficiency coefficient by the recursive least squares method based on the steady-state time sequence, and calculates the sealing performance index in combination with the door disturbance temperature characteristics; and a performance evaluation module, which calculates the temperature fluctuation degree, determines the equipment performance grade in multiple dimensions and outputs a comprehensive detection report. The present application can synchronously detect the indicators such as the insulation of the freezer, the door seal and the temperature uniformity, and solves the problem of integrated on-line detection of multiple performances.

Description

Technical Field

[0001] This invention relates to the field of online performance testing technology for medical ultra-low temperature freezers, and specifically to a comprehensive performance testing system for medical ultra-low temperature freezers. Background Technology

[0002] Medical ultra-low temperature freezers are critical equipment for preserving biological samples, vaccines, and biological products. Their internal temperatures often need to be maintained at -80°C or even lower, placing stringent requirements on thermal insulation, door sealing, and temperature field uniformity. Currently, the thermal insulation performance of ultra-low temperature freezers is typically tested using standard methods. This requires placing the freezer in a specified constant-temperature environment after power is off, recording the temperature rise curve using multiple temperature probes inside the freezer, and evaluating the insulation capacity based on the time it takes for the temperature to rise from a certain low point to a specific value. This method not only requires specialized testing laboratories and a long stabilization process, but more importantly, it necessitates emptying the freezer and interrupting normal use. It cannot reflect the thermal characteristics of the equipment under actual load operation and is difficult to detect performance degradation caused by vacuum insulation layer degradation or door seal aging during use. Testing refrigeration efficiency requires invasive access to pressure or flow measurement devices, which is complex and cannot be performed simultaneously with insulation performance testing. Door seal sealing performance testing during routine maintenance often relies on manual visual inspection of the door seal for deformation or frost, or methods such as dragging paper clips or measuring pressure drop by inflating the cabinet after closing the door. These methods are highly subjective, lack quantification, and cannot be implemented during daily operation where samples are stored inside the cabinet, thus failing to provide continuous quantitative indicators of sealing performance degradation. Similarly, evaluating temperature uniformity and fluctuation typically requires arranging numerous thermocouples in a standard grid under no-load conditions, which are then removed after the test, and this also lacks linearity.

[0003] In recent years, some online monitoring solutions for freezers have been developed to collect and upload temperature data in real time by installing temperature sensors and communication modules, triggering alarms when temperatures exceed limits. However, these solutions only reflect whether the temperature in certain locations within the freezer exceeds preset ranges, failing to identify the root cause of temperature anomalies as insulation degradation, reduced cooling capacity, or door seal failure. They also struggle to provide quantitative parameters characterizing the freezer's physical properties, such as thermal resistance and heat capacity. Some solutions attempt to analyze and process temperature data, but when individual probes in a multi-point sensor system experience zero-point drift or malfunction, the lack of effective data verification and rejection mechanisms easily introduces erroneous temperature values ​​into subsequent calculations, leading to misjudgments. Furthermore, the disturbance to the internal temperature field caused by door opening and closing events and its recovery process contain rich information about the freezer's sealing performance and cold storage capacity. Existing detection and monitoring systems do not utilize this process to quantitatively extract sealing performance and recovery capabilities, resulting in information waste. Since ultra-low temperature freezers often require continuous operation year-round in actual use, frequent shutdowns for testing are not feasible. Traditional periodic performance evaluations can only extend the cycle, during which the freezer's performance may significantly degrade without being detected, posing a potential risk to sample safety. Therefore, there is an urgent need in this field for a comprehensive testing system that can automatically and synchronously extract multiple key performance indicators such as thermal resistance, refrigeration efficiency, sealing performance, temperature uniformity, and temperature fluctuation from daily operating data without interrupting the normal use of the freezer or requiring a dedicated testing environment, so as to achieve online assessment of the freezer's health status and early warning of degradation. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this invention is to provide a comprehensive performance testing system for medical ultra-low temperature freezers, which solves the problem of difficulty in online comprehensive testing of the insulation performance, door seal sealing performance, and temperature uniformity of ultra-low temperature freezers during operation. This invention simultaneously collects multiple temperature points inside the freezer, compressor power, and door opening / closing signals. After outlier removal and low-pass filtering, it automatically divides the system into steady-state insulation, transient cooling, and door opening disturbance periods based on door status and temperature change rate. During the steady-state period, a recursive least squares algorithm with a forgetting factor is used to solve the lumped-parameter thermal balance equation online, identifying the equivalent thermal resistance, heat capacity, and refrigeration efficiency coefficient. During the door opening disturbance period, the peak temperature rise and recovery time are extracted, and the sealing performance index is calculated in conjunction with thermal characteristic parameters. Simultaneously, the standard deviation and maximum deviation of the effective sensor temperatures are calculated to generate temperature fluctuation and uniformity signals. Finally, a comprehensive test report is output, achieving quantitative assessment of performance degradation without interrupting freezer operation.

[0005] This invention provides a comprehensive performance testing system for medical ultra-low temperature freezers, comprising:

[0006] The sensor acquisition module acquires synchronous timing data stream signals;

[0007] The data preprocessing module receives synchronous time-series data stream signals, performs Grubbs criterion anomaly removal and interpolation on multi-point temperature signals, and performs low-pass filtering on compressor electrical parameter signals to form a preprocessed dataset signal.

[0008] The state classification module receives the preprocessed dataset signal and, based on the door opening / closing state change edge and the comparison results of the temperature change rate with the threshold, divides the operating period into steady-state heat preservation, transient cooling, and door opening disturbance recovery period signals.

[0009] The parameter identification module receives signals during steady-state insulation and door opening disturbance recovery periods. Under the steady-state insulation period signal, it solves the heat balance equation by recursive least squares to generate thermal characteristic parameter signals such as equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient. Under the door opening disturbance recovery period signal, it extracts the door closing temperature, peak rise, and recovery time, and combines the thermal characteristic parameters to generate a sealing performance index signal.

[0010] The performance evaluation module receives thermal characteristic parameter signals, sealing performance index signals, and preprocessed dataset signals, calculates the effective temperature standard deviation to generate temperature fluctuation signals, and summarizes and outputs a comprehensive test report signal.

[0011] In one embodiment of the present invention, after receiving the synchronous timing data stream signal, the data preprocessing module performs outlier detection based on the Grubbs criterion on each temperature data sequence in the multi-point temperature signal, marks the data points determined to be outliers and removes them from the sequence, and fills the positions of the removed data points by linear interpolation of adjacent valid data points to form a continuous and complete temperature data sequence; the data preprocessing module performs low-pass filtering and smoothing on the average power signal in the compressor electrical parameter signal to remove high-frequency noise components, and combines the processed multi-point temperature data sequence with the smoothed compressor average power signal, door switch status signal and ambient temperature signal to form a preprocessed dataset signal.

[0012] In one embodiment of the present invention, before performing outlier removal based on the Grubbs criterion on multi-point temperature signals, the data preprocessing module first calculates the arithmetic mean and standard deviation of all temperature sensor readings at the current moment. Using the standard deviation as a benchmark, it determines the critical coefficient corresponding to the number of sensors in the Grubbs criterion. The statistic obtained by dividing the absolute value of the deviation between each sensor reading and the arithmetic mean by the standard deviation is compared with the critical coefficient. Sensor readings exceeding the critical coefficient are judged as outliers and removed, while sensor readings not exceeding the critical coefficient are retained as valid data. After completing outlier removal, the data preprocessing module renumbers and counts the retained valid sensors, using the number of valid sensors as an input parameter for subsequent state classification and performance evaluation. When the number of valid sensors is lower than the preset minimum number of sensors, a sensor fault alarm signal is generated. The calculation process of comparing the statistic obtained by dividing the absolute value of the deviation between each sensor reading and the arithmetic mean by the standard deviation with the critical coefficient is as follows:

[0013] ;

[0014] in, Let be the Grubbs statistic corresponding to the i-th temperature sensor at the k-th sampling time. Here, N represents the raw temperature reading of the i-th sensor at the k-th sampling time, and N is the total number of temperature sensors connected at the current time. This is the summation of all sensor temperature readings at the current moment. This represents the sample standard deviation of all temperature readings at the current moment. The arithmetic mean of the current temperature. This represents the absolute deviation between the temperature of a single sensor and its mean value.

[0015] In one embodiment of the present invention, after receiving the preprocessed dataset signal, the state classification module monitors the logic level change edge of the door switch state signal in real time. When it detects that the door switch state changes from the closed level to the open level, it records the opening time; when it detects that the door switch state changes from the open level to the closed level, it records the closing time and marks a continuous time period after the closing time as the door opening disturbance recovery period signal. Within the interval where the door switch state signal remains at the closed level, the state classification module performs frame-by-frame differential operation on the equivalent average temperature obtained by fusing the multi-point temperature signals after consistency verification, calculates the temperature change rate per unit time, and marks the corresponding time period as the transient cooling period signal when the absolute value of the temperature change rate is greater than or equal to the preset cooling change rate threshold and the equivalent average temperature is higher than the set target temperature; when the absolute value of the temperature change rate is less than the preset heat preservation change rate threshold and the equivalent average temperature fluctuates near the set target temperature, the corresponding time period is marked as the steady-state heat preservation period signal.

[0016] In one embodiment of the present invention, when the state classification module divides the steady-state heat preservation period signal, it needs to simultaneously meet the following conditions: the door switch status signal remains at the closed level, the absolute value of the difference between the equivalent average temperature and the set target temperature is less than the preset temperature deviation allowable value, and the standard deviation of the equivalent average temperature in multiple consecutive sampling periods is less than the preset temperature fluctuation allowable value. The continuous time period in which all three conditions are met is marked as the effective steady-state heat preservation period signal. The state classification module outputs the start time and end time of the effective steady-state heat preservation period signal to the parameter identification module, and identifies the start and stop status of the compressor electrical parameter signal within the effective steady-state heat preservation period. When the average power of the compressor is greater than the preset start power threshold, it is determined that the compressor is in the running state. When the average power of the compressor is less than the preset standby power threshold, it is determined that the compressor is in the stopped state. The compressor start and stop status sequence signal is generated as a component of the steady-state heat preservation period signal.

[0017] In one embodiment of the present invention, after receiving the steady-state heat preservation period signal, the parameter identification module extracts the equivalent average temperature sequence, ambient temperature sequence, and compressor average power sequence corresponding to the steady-state heat preservation period from the preprocessed dataset signal. The parameter identification module establishes a lumped parameter heat balance equation, approximating the derivative of the equivalent average temperature with respect to time by the ratio of the temperature difference between adjacent sampling times to the sampling interval, using the difference between the ambient temperature and the equivalent average temperature as the heat transfer driving force term, and using the product of the compressor average power and the compressor start-stop state sequence signal as the refrigeration input term, thus constructing a discrete-time difference equation containing three unknown parameters: equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient. The parameter identification module uses a recursive least squares algorithm with a forgetting factor to perform online parameter estimation of the discrete-time difference equation. The value of the forgetting factor gives the algorithm greater weight to recent data. The parameter estimation value is updated recursively by sampling point. When the relative change of the parameter estimation value is less than the preset convergence threshold for multiple consecutive sampling periods, the converged equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient are output as thermal characteristic parameter signals.

[0018] In one embodiment of the present invention, the recursive least squares algorithm with a forgetting factor used in the parameter identification module includes the following recursive steps: At each sampling moment, a regression vector is constructed based on the current equivalent average temperature, the previous equivalent average temperature, the ambient temperature, the compressor average power, and the compressor start / stop status. The predicted temperature value at the current moment is calculated using the parameter estimate value at the previous moment. The difference between the actual equivalent average temperature at the current moment and the predicted temperature value is used as the prediction error. The gain vector at the current moment is calculated by combining the forgetting factor and the covariance matrix at the previous moment. The parameter estimate value is corrected and updated using the gain vector and the prediction error, and the covariance matrix is ​​updated simultaneously. When the change amplitude of the parameter estimate value at multiple consecutive moments during the recursive process is lower than the preset convergence threshold, the current parameter estimate value is locked as the final equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient, and stored in a fixed manner until the parameter identification is triggered again in the next detection cycle. The formula for calculating the gain vector is as follows:

[0019] ;

[0020] in, This is the recursive least squares gain vector with forgetting factor at the k-th sampling time. Let be the parameter covariance matrix at the (k-1)th sampling time. The thermal equilibrium regression vector is constructed based on the filtered effective temperature sequence at the k-th sampling time. The preset forgetting factor value is 0 < <1, Let be the transpose of the regression vector. It is the scalar term of the product of the regression vector, covariance matrix, and transpose vector.

[0021] In one embodiment of the present invention, after receiving the door opening disturbance recovery period signal, the parameter identification module extracts the equivalent average temperature at the door closing time, the equivalent average temperature sequence during the door opening disturbance recovery period, and the compressor average power sequence from the preprocessed dataset signal; the parameter identification module traverses the equivalent average temperature sequence during the door opening disturbance recovery period and finds the temperature maximum value during the temperature recovery process as the temperature recovery peak value; the parameter identification module calculates the difference between the temperature recovery peak value and the set target temperature as the temperature recovery amplitude, and calculates the time required for the equivalent average temperature to fall back to the sum of the set target temperature and the preset recovery deviation value from the time corresponding to the temperature recovery peak value as the door opening recovery time; the parameter identification module calculates the theoretical thermal time constant as the product of the equivalent thermal resistance and the equivalent heat capacity based on the identified equivalent thermal resistance and equivalent heat capacity, uses the ratio of the theoretical thermal time constant to the door opening recovery time as the reference recovery rate, and weights and combines the reference recovery rate with the inverse change rate of the equivalent thermal resistance relative to the factory reference thermal resistance to generate a sealing performance index signal.

[0022] In one embodiment of the present invention, when generating the sealing performance index signal, the parameter identification module introduces the factory reference thermal resistance and factory reference recovery time as reference benchmarks. The factory reference thermal resistance and factory reference recovery time are pre-stored in the system's non-volatile memory and are identified and recorded by the parameter identification module during factory testing of the same model of refrigerator under standard operating conditions using the same recursive least squares algorithm as online testing. The parameter identification module uses the ratio of the equivalent thermal resistance obtained online to the factory reference thermal resistance as the thermal resistance degradation factor, and the ratio of the door opening recovery time measured online to the factory reference recovery time as the recovery time degradation factor. The reciprocal of the thermal resistance degradation factor and the reciprocal of the recovery time degradation factor are weighted and summed. The weights are allocated according to the sensitivity differences of the thermal resistance degradation factor and the recovery time degradation factor at different aging stages of the refrigerator. The weighted summation result is the sealing performance index signal. The calculation formula for the sealing performance index signal is as follows:

[0023] ;

[0024] Where S represents the real-time sealing performance index of the medical ultra-low temperature freezer. These are the weighting coefficients corresponding to the thermal resistance degradation factor. The weighting coefficients corresponding to the recovery time degradation factor are used. R is the factory-standard equivalent thermal resistance, and R is the real-time equivalent thermal resistance identified by the recursive least squares gain vector algorithm with forgetting factor in the second band. The door opening and recovery time is the factory standard. The measured door opening recovery time is the time taken during the period of door opening disturbance. This is the fundamental weighting constant for thermal resistance sensitivity. This is the thermal resistance aging degradation adjustment coefficient. This is an exponential correction term that varies with the degree of thermal resistance degradation. =1 is the weight normalization constraint.

[0025] In one embodiment of the present invention, the performance evaluation module receives a preprocessed dataset signal, extracts the effective sensor temperature data corresponding to the steady-state heat preservation period signal, calculates the standard deviation of the temperature values ​​of all effective sensors at the same sampling time, uses the standard deviation as the instantaneous value of temperature uniformity at that sampling time, and performs an arithmetic mean of the instantaneous values ​​of temperature uniformity at all sampling times within the steady-state heat preservation period to generate a temperature uniformity signal; the performance evaluation module calculates the standard deviation of the time series of equivalent average temperature within the steady-state heat preservation period to generate a temperature fluctuation signal; the performance evaluation module compares the equivalent thermal resistance signal with a preset insulation performance grading threshold to determine the insulation performance level, compares the refrigeration efficiency coefficient signal with a preset refrigeration efficiency grading threshold to determine the refrigeration efficiency level, compares the sealing performance index signal with a preset sealing performance grading threshold to determine the sealing performance level, compares the temperature fluctuation signal with a preset fluctuation grading threshold to determine the temperature stability level, and compares the temperature uniformity signal with a preset uniformity grading threshold to determine the temperature uniformity level, and summarizes the insulation performance level, refrigeration efficiency level, sealing performance level, temperature stability level, and temperature uniformity level to form a comprehensive test report signal.

[0026] Beneficial Effects: This invention provides a comprehensive performance testing system for medical ultra-low temperature freezers. By simultaneously collecting multiple temperature points within the freezer, compressor power, and door opening / closing signals, and after outlier removal and low-pass filtering, the system automatically divides the timeframes into steady-state insulation, transient cooling, and door opening disturbance periods based on door status and temperature change rate. During the steady-state period, a recursive least squares algorithm with a forgetting factor is used to solve the lumped-parameter thermal balance equation online, identifying the equivalent thermal resistance, heat capacity, and refrigeration efficiency coefficient. During the door opening disturbance period, the system extracts the temperature rise peak and recovery time, and calculates the sealing performance index based on thermal characteristic parameters. Simultaneously, the system calculates the standard deviation and maximum deviation of the effective sensor temperatures, generating temperature fluctuation and uniformity signals. Finally, a comprehensive test report is output, enabling quantitative assessment of performance degradation without interrupting freezer operation. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a system architecture diagram of a comprehensive performance testing system for medical ultra-low temperature freezers;

[0029] Figure 2 This is a flowchart of the internal processing of the data preprocessing module;

[0030] Figure 3A flowchart for identifying recursive least squares parameters during the steady-state insulation period. Detailed Implementation

[0031] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0032] Please see Figures 1-3 The diagram illustrates a comprehensive performance testing system for a medical ultra-low temperature freezer according to the present invention. The system includes a sensing and acquisition module for acquiring synchronous time-series data stream signals; a data preprocessing module for receiving the synchronous time-series data stream signals, performing Grubbs' criterion anomaly removal and interpolation on multi-point temperature signals, and low-pass filtering the compressor electrical parameter signals to form a preprocessed dataset signal; a state classification module for receiving the preprocessed dataset signal, and dividing the operating period into steady-state insulation, transient cooling, and door opening disturbance recovery periods based on the door opening / closing state change edges and the comparison results of the temperature change rate with a threshold; and parameter identification... The identification module receives signals from steady-state insulation and door opening disturbance recovery periods. Under the steady-state insulation period signal, it solves the heat balance equation through recursive least squares to generate thermal characteristic parameter signals such as equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient. Under the door opening disturbance recovery period signal, it extracts the door closing temperature, peak rise, and recovery time, and combines the thermal characteristic parameters to generate a sealing performance index signal. The performance evaluation module receives the thermal characteristic parameter signals, sealing performance index signals, and preprocessed dataset signals, calculates the effective temperature standard deviation to generate a temperature fluctuation signal, and summarizes and outputs a comprehensive test report signal.

[0033] Figure 1As shown, this process is the complete top-level working link of the entire testing system, connecting five basic functional modules. It also incorporates multi-condition branching logic and data feedback loop. The starting unit of the process is the sensor acquisition module, which collects synchronous time-series data streams. This module is equipped with multiple temperature sensor probes, compressor electrical acquisition devices, door switch level detectors, and ambient temperature sensors. All acquisition devices synchronously trigger sampling operations and output synchronous time-series raw data streams with unified timestamps. The raw data includes four types of basic sensor information: multi-point temperature inside the cabinet, real-time operating power of the compressor, door opening and closing status, and external ambient temperature. It is the only source of raw data for all performance calculations of the entire machine. After acquisition, the data is directly transmitted to the next level functional unit. The second unit is the data preprocessing module, which removes anomalies and filters the output dataset. This module performs a complete set of processing logic, including temperature data Grubbs anomaly verification, anomaly point interpolation completion, compressor power low-pass noise reduction, valid sensor quantity verification, and multi-type signal time sequence splicing. It filters noise and failure values ​​in the original sensor data and outputs a standardized preprocessed dataset with time sequence alignment and no abnormal fluctuations. This eliminates the interference of underlying data defects on upper-level operating condition division, parameter identification, and performance evaluation. The processed dataset is then uniformly sent to the corresponding unit of the state classification module. The third unit is the state classification module, which divides the operating segments into three categories. This module reads the door switch level signal and equivalent average temperature time sequence from the preprocessed dataset, continuously monitors the door level jump moments and records the opening and closing time nodes, calculates the cabinet temperature change rate frame by frame within the cabinet door closing interval, and completes the distinction of operating conditions by combining temperature deviation and temperature fluctuation threshold, dividing the signal into three independent time sequence signals: transient cooling, steady-state insulation, and door opening disturbance recovery. After the division is completed, the signal is transferred to the time segment type determination branch unit. This unit matches each continuous sampling frame with the corresponding operating condition type. There are no interrogative judgment statements. The signal is divided into three independent data flow branches according to the three operating conditions.

[0034] The first branch corresponds to the transient cooling period, which only caches data without identification units. This unit only temporarily stores the complete time-series data of the cooling interval without performing any parameter solving or performance calculations. After storage, it is re-sent to the state classification module through the data backflow loop to divide the three types of runtime units and continuously monitor the state switching status after the cooling condition ends. The second branch corresponds to the steady-state insulation period, which sends data to the RLS parameter identification unit. This unit extracts the complete time-series data of the steady-state interval and starts a recursive least squares iterative operation with a forgetting factor to solve for the cabinet's equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient thermal characteristic parameters. The third branch corresponds to the door opening disturbance recovery period, which extracts the temperature peak / recovery time unit. This unit traverses the temperature rise sequence after the door is closed to locate the temperature peak, calculates the door opening recovery time required for the cabinet temperature to fall back to the target range, and records all temperature change characteristic parameters under the door disturbance condition. The data from the two branches involved in parameter calculation are uniformly aggregated to the thermal characteristic parameter + sealing index unit output by the parameter identification module. This unit integrates the thermal characteristic parameters identified under steady-state conditions with the temperature feature data extracted under door disturbance conditions, and generates a sealing performance index signal by adaptive weighted calculation based on the factory baseline parameters. It simultaneously outputs both the thermal characteristic parameter signal and the sealing performance index signal, both of which are sent to the corresponding unit in the performance evaluation module. The performance evaluation module calculates temperature uniformity / fluctuation, determining each performance unit as the core unit for overall system performance quantification. This unit retrieves effective temperature measurement data from the steady-state range within the preprocessed dataset, calculates the instantaneous value of temperature uniformity, and obtains the mean value to obtain the temperature uniformity signal. Based on the steady-state equivalent average temperature time series, it calculates the standard deviation to obtain the temperature fluctuation signal. Then, it compares the thermal characteristic parameters, sealing index, temperature uniformity, and temperature fluctuation with preset grading thresholds, sequentially determining the five performance levels: insulation performance, cooling efficiency, sealing performance, temperature stability, and temperature uniformity. This completes the full-dimensional performance quantification and grading. All grading results are then aggregated and transferred to... The terminal output integrated test report signal unit integrates five performance levels, thermal characteristic parameters, sealing index, temperature fluctuation and uniformity values ​​to generate a standardized integrated test report signal for the entire machine. It can be directly output to the host computer to complete data storage, report printing and equipment fault early warning. The entire process realizes a closed-loop operation from raw sensor acquisition, data cleaning, working condition differentiation, multi-branch parallel parameter calculation to comprehensive performance rating. The data feedback loop of the transient cooling branch realizes uninterrupted working condition monitoring throughout the entire cycle. Data of each functional unit is interconnected, fully covering the full-dimensional performance testing needs of medical ultra-low temperature freezers in terms of temperature, insulation, sealing and refrigeration.

[0035] Specifically, a complete sensor network is deployed at the front end of the system. Thermocouples or thermistors are installed at multiple representative spatial locations inside the freezer. These locations cover the geometric center of the freezer, the center points of each shelf, and the edge areas near the door seal and walls, ensuring that the collected multi-point temperature signals fully reflect the spatial distribution characteristics of the temperature field inside the freezer. Voltage and current transformers are connected to the compressor power supply line. The average power signal reflecting the real-time power consumption of the compressor is obtained through multiplication and low-pass filtering, thus characterizing the operating intensity of the refrigeration system. A door open / closed status sensor installed at the junction of the door frame and the freezer body outputs a logic level in real time to indicate the open / closed status of the door. An ambient temperature sensor is installed on the outside of the freezer in a location not directly affected by the heat dissipation of the freezer body, providing an environmental benchmark for subsequent thermal balance analysis. The analog signals output by all sensors are sampled by a synchronous analog-to-digital converter driven by the same clock and assigned a unified timestamp. This ensures that the digital signals of multi-point temperature, compressor voltage and current, door switch status, and ambient temperature are precisely aligned and packaged into a synchronous timing data stream signal for transmission to downstream modules. This synchronization alignment mechanism is the physical basis for all subsequent time-related analyses.

[0036] After receiving the synchronous time-series data stream signal, the data preprocessing module performs two key data cleaning operations. For multi-point temperature signals, the module first calculates the arithmetic mean and standard deviation of all temperature sensor readings at the current moment. Using the standard deviation as a benchmark, it determines the critical coefficient in the Grubbs criterion corresponding to the current number of valid sensors. The absolute value of the deviation of each sensor reading from the arithmetic mean is divided by the standard deviation to obtain a statistic, which is then compared with the critical coefficient. Sensor readings exceeding the critical coefficient are judged as outliers and removed from the sequence. Linear interpolation of adjacent valid data points is used to fill in the removed positions, thus forming a continuous and complete temperature data sequence. This mechanism fundamentally solves the problem of distorted subsequent performance evaluation caused by erroneous data introduced by some sensors due to drift, poor contact, or malfunction. When the number of valid sensors remaining after removing outliers is lower than the preset minimum number, the system generates a sensor fault alarm signal to prompt maintenance. For compressor electrical parameter signals, the data preprocessing module performs low-pass filtering and smoothing on the average power signal to remove high-frequency noise components. Finally, the processed multi-point temperature data sequence, the smoothed compressor power sequence, the door switch status sequence, and the ambient temperature sequence are combined into a preprocessed dataset signal. The statistic obtained by dividing the absolute value of the deviation of each sensor reading from the arithmetic mean by the standard deviation is compared with the critical coefficient. The calculation process is as follows:

[0037] ;

[0038] in, Let be the Grubbs statistic corresponding to the i-th temperature sensor at the k-th sampling time. Here, N represents the raw temperature reading of the i-th sensor at the k-th sampling time, and N is the total number of temperature sensors connected at the current time. This is the summation of all sensor temperature readings at the current moment. This represents the sample standard deviation of all temperature readings at the current moment. The arithmetic mean of the current temperature. This represents the absolute deviation between the temperature of a single sensor and its mean. After each synchronized time-series data acquisition, the system iterates through all temperature measurement channels, substituting each value into the formula to calculate the corresponding Grubbs statistic. The statistic is then compared to the Grubbs critical coefficient for the number of matching sensors. Temperature sampling points exceeding the critical coefficient are identified as abnormal noise points and directly removed. The resulting data gaps are filled by linear interpolation of adjacent effective temperatures, forming a continuous, uninterrupted temperature time-series sequence. After anomaly cleaning, the system recounts the number of effective sensors. When the number of effective sensors falls below a preset minimum sensor count threshold, a sensor fault alarm signal is output. This formula is the foundational computational unit for the entire detection algorithm. All subsequent state classifications, recursive least squares parameter identification, and sealing performance index calculations rely on the clean and effective temperature dataset output by this formula. Traditional 3σ anomaly detection methods cannot adapt to small-sample detection scenarios with a limited number of temperature measurement channels in medical refrigerators. The Grubbs statistic achieves accurate anomaly screening through a critical coefficient bound to the number of sensors, eliminating the judgment bias caused by manually setting fixed thresholds. It can adapt to 24 / 7 unattended continuous detection conditions, effectively avoiding problems such as equivalent average temperature calculation deviation and incorrect division of steady-state insulation periods caused by abnormal temperature data. From the data source level, it ensures the iterative convergence of the subsequent recursive least squares algorithm with forgetting factor. If the data purification step of this formula is missing, jump noise will cause the regression vector value to fluctuate drastically, causing the identification results of equivalent thermal resistance and equivalent heat capacity to deviate from the actual thermal characteristics of the equipment, further interfering with the quantitative accuracy of the sealing performance index. In the mandatory testing scenario of medical low-temperature equipment required by drug regulatory authorities, this standardized statistical calculation formula ensures the traceability of temperature data and the reproducibility of test results, providing reliable raw calculation input for the entire comprehensive performance testing system.

[0039] like Figure 2As shown, the entire data preprocessing workflow uses synchronous time-series sensor data as the initial input. The first execution unit of the workflow receives the synchronous time-series data stream signal. This unit is responsible for receiving multi-channel parallel time-series sensor data transmitted from the front-end acquisition device. The data synchronously carries four different dimensions of raw signals: temperature, electrical, door status, and ambient temperature. All signals carry the same sampling timestamp, ensuring the basic conditions for subsequent multi-dimensional data alignment and calculation. The raw data has not undergone noise reduction and anomaly filtering, and contains abrupt changes in values ​​caused by sensor instantaneous drift and electromagnetic interference, making it impossible to directly send it to the subsequent calculation unit for analysis. The next execution unit splits the multi-point temperature, compressor power, door status, and ambient temperature signals. This unit performs data splitting and storage according to the signal data type, aggregating all temperature measurement channel data into an independent temperature dataset, dividing the compressor operating average power into a separate electrical parameter dataset, and generating independent signal queues for door switch high and low level signals and ambient temperature sensor values. The splitting operation retains the unified sampling timestamp corresponding to each group of data to avoid timing misalignment between different types of data, providing data partitioning for independent channel-specific calculations. After data splitting is completed, the individual sensor is used to calculate the Grubbs statistic. This unit performs statistical calculations on each segmented temperature sensor data individually. It calculates standardized statistics based on the mean temperature of all channels and the sample standard deviation. The calculation process covers all connected temperature sensors. After a single channel's calculation is complete, it enters the branch decision unit. This unit only compares the Grubbs statistic value with a preset critical coefficient, without any interrogative statements. If the statistic value exceeds the critical coefficient, the current channel's data is considered abnormal, and the data is transferred to the outlier removal and linear interpolation temperature sequence completion unit. This unit marks the abnormal data points and directly removes the corresponding values. It then extracts the adjacent valid temperature samples before and after the outlier point and uses a linear interpolation algorithm to generate replacement values ​​to fill the missing points, forming a continuous temperature time series without breaks. After single-channel anomaly handling is complete, it returns to the single sensor via a loop to calculate the Grubbs statistic. The unit continuously traverses the remaining unprocessed temperature measurement channels until all temperature channels have completed anomaly verification and repair.

[0040] After all temperature measurement channels are processed, the data flows to the unit that counts the total number of valid sensors (N). This unit recounts the total number of working temperature measurement sensors after removing abnormal channels, records the number of valid sensors as the basic input parameter for subsequent system-wide calculations, and then proceeds to the second-level branch judgment unit. This unit compares the total number of valid sensors with the system's preset minimum number of usable sensors. If the number of valid sensors does not meet the minimum standard, the process flows to the output sensor fault alarm signal unit. This unit pushes out a standardized fault indication signal, records the failed sensor number and the failure sampling time, and after completing the alarm output, it flows to the power low-pass filter and splices the data to generate a preprocessed dataset signal unit. If the number of valid sensors meets the minimum usage standard, the process jumps directly to this final unit. The power low-pass filtering and splicing preprocessed dataset signal unit consists of two steps. The first step performs low-pass filtering and smoothing on the average power data of the compressor after the power grid is split, filtering out high-frequency noise caused by power grid fluctuations to obtain a stable and continuous compressor power sequence. The second step splices and integrates the repaired complete temperature sequence, the filtered compressor power sequence, the door switch status level sequence, and the ambient temperature time series data according to a unified timestamp to generate a preprocessed dataset signal with a unified format, no abnormal noise, and time alignment. This signal is then sent to the next functional module of the system for time period segmentation calculation. The entire process includes a two-layer branch judgment logic and a sensor traversal loop link, fully covering all processing steps of the original sensor data from reception, splitting, anomaly cleaning, fault verification to noise reduction and integration. The data before and after each execution unit are closely connected, eliminating the deviation caused by noise and faulty data on the subsequent overall system performance test results from the underlying data source level, ensuring the stability and reliability of the entire system's test data.

[0041] The state classification module is responsible for automatically parsing the continuous operation of the ultra-low temperature freezer into time periods with different physical meanings. This parsing is a prerequisite for implementing the divide-and-conquer strategy. The module monitors the logic level changes of the door opening and closing status signals in real time. When a transition from the closed level to the open level is detected, the door opening time is recorded. When a transition from the open level to the closed level is detected, the door closing time is recorded. The continuous time period following the door closing time is marked as the door opening disturbance recovery period signal. Within the range where the door open / closed status signal remains at the closed level, the module performs frame-by-frame differential calculations on the equivalent average temperature obtained by fusing multi-point temperature signals after consistency verification to calculate the rate of temperature change per unit time. When the absolute value of the rate of temperature change is greater than or equal to a preset cooling rate threshold and the equivalent average temperature is higher than the set target temperature, the corresponding time period is marked as a transient cooling period signal. When the absolute value of the rate of temperature change is less than a preset insulation rate threshold and the absolute value of the difference between the equivalent average temperature and the set target temperature is less than a preset temperature deviation allowable value, and the standard deviation of the equivalent average temperature over multiple consecutive sampling periods is less than a preset temperature fluctuation allowable value, the corresponding time period is marked as a steady-state insulation period signal. The joint judgment of these three conditions ensures that the time period marked as steady-state insulation truly reflects the thermal equilibrium state where the cabinet door is closed and the temperature inside the cabinet is sufficiently stable near the target value. The status classification module will also identify the start / stop status of the compressor average power during the steady-state insulation period, and generate a compressor start / stop status sequence signal by comparing it with the start power threshold and the standby power threshold, which is output as a component of the steady-state insulation period signal.

[0042] The parameter identification module performs online identification of thermal characteristic parameters under signal triggering during the steady-state insulation period. This is the core component of the system for non-invasive insulation and refrigeration performance evaluation. The module extracts the equivalent average temperature sequence, ambient temperature sequence, compressor average power sequence, and compressor start-stop state sequence corresponding to the steady-state insulation period from the preprocessed dataset signal, establishing a lumped-parameter heat balance equation. This equation approximates the derivative of the equivalent average temperature with respect to time using the ratio of the temperature difference between adjacent sampling times to the sampling interval. It uses the difference between the ambient temperature and the equivalent average temperature as the heat transfer driving force term and the product of the compressor average power and the compressor start-stop state as the refrigeration input term, thus forming a discrete-time difference equation containing three unknown parameters: equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient. These three parameters respectively characterize the degree of heat leakage from the cabinet to the environment, the cold storage capacity of the cabinet, and the efficiency of the refrigeration system in converting electrical power into effective refrigeration capacity. The parameter identification module employs a recursive least squares algorithm with a forgetting factor to solve the equation online. The forgetting factor assigns greater weight to recently acquired data, enabling continuous tracking of the slow changes in the refrigerator's thermal characteristics over long-term use. Each step of the recursive process utilizes the parameter estimates from the previous moment, the current and previous equivalent average temperatures, ambient temperature, and compressor power to construct a regression vector and calculate the predicted temperature. The deviation between the predicted and actual values ​​serves as the error signal driving parameter updates. Point-by-point correction of the three unknown parameters is achieved through synchronous updates of the gain vector and covariance matrix. When the relative change in the parameter estimates is less than a preset convergence threshold for multiple consecutive sampling periods, the algorithm is considered converged. The current equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient are locked and stored as thermal characteristic parameter signals, outputting until the next detection period triggers identification again.

[0043] The parameter identification module performs a quantitative evaluation of sealing performance upon triggering a signal during the door opening disturbance recovery period. The module extracts the equivalent average temperature at the moment of door closing, iterates through the equivalent average temperature sequence during the door opening disturbance recovery period to find the temperature maximum during the temperature rise process as the temperature rise peak, calculates the difference between the temperature rise peak and the set target temperature as the temperature rise amplitude, and calculates the time required for the equivalent average temperature to fall back to the sum of the set target temperature and the preset recovery deviation value, starting from the moment corresponding to the temperature rise peak, as the door opening recovery time. This recovery process comprehensively reflects the amount of hot and humid air intrusion during door opening and the refrigeration system's ability to bring the cabinet back to the target temperature after door closing. The module introduces the factory reference thermal resistance and factory reference recovery time of the same model of refrigerator, pre-stored in the system's non-volatile memory, as references. These two reference values ​​are identified and recorded under standard operating conditions during the refrigerator's factory testing using the same recursive least squares algorithm as this system. The module defines the ratio of the equivalent thermal resistance obtained online to the factory baseline thermal resistance as the thermal resistance degradation factor, and the ratio of the door opening recovery time measured online to the factory baseline recovery time as the recovery time degradation factor. Since the equivalent thermal resistance degradation directly reflects the deterioration of the cabinet's insulation layer, and the extension of the recovery time is closely related to the sealing degree of the door seal in addition to the influence of thermal resistance degradation, the module performs a weighted sum of the reciprocals of the thermal resistance degradation factor and the recovery time degradation factor. The weights are allocated according to the sensitivity differences of the two factors at different aging stages, and the weighted sum results in a sealing performance index signal. This index separates the degradation of the door seal's sealing performance from the overall thermal resistance degradation effect, enabling independent quantitative evaluation of sealing performance without disassembling the door seal. The formula for calculating the gain vector is as follows:

[0044] ;

[0045] in, This is the recursive least squares gain vector with forgetting factor at the k-th sampling time. Let be the parameter covariance matrix at the (k-1)th sampling time. The thermal equilibrium regression vector is constructed based on the filtered effective temperature sequence at the k-th sampling time. The preset forgetting factor value is 0 < <1, Let be the transpose of the regression vector. The product of the regression vector, covariance matrix, and transpose vector is a scalar term. In each sampling period, the algorithm first constructs a regression vector based on the preprocessed dataset, then uses the converged thermal characteristic parameter estimates from the previous time step to calculate the theoretical temperature prediction. The difference between the actual equivalent average temperature and the prediction value is used to obtain the prediction error. The prediction error, forgetting factor, and historical covariance matrix are then substituted into the gain vector calculation formula. The gain vector value determines the correction magnitude of the current sampled data to the parameter estimates; a larger gain vector value indicates higher reliability of the current temperature and power data, and a larger parameter iteration correction magnitude. After obtaining the gain vector, the three sets of unknown parameters—equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient—are updated synchronously. Simultaneously, the covariance matrix is ​​iteratively updated to record the parameter estimation uncertainty. This process is continuously iterated until the relative change in the parameter estimates over multiple consecutive periods is less than the convergence threshold, at which point the final thermal characteristic parameters are locked and stored. This formula relies entirely on the pure temperature time-series data output by the preceding Grubbs statistic for calculation. Abnormal data loss will directly cause regression vector distortion, and the gain vector will continue to oscillate and fail to converge. The final identified equivalent thermal resistance will have significant errors. Since the equivalent thermal resistance is the core input variable of the sealing performance index formula, the gain vector calculation formula builds the core calculation bridge between temperature preprocessing and sealing performance evaluation. The forgetting factor adapts to the slow aging of equipment thermal characteristics by reducing the weight of historical old data. Unlike the traditional least squares algorithm without forgetting factor, which is prone to parameter results being solidified by early historical data, this formula can simultaneously meet the testing needs of multiple scenarios such as factory calibration, annual inspection, and online fault diagnosis. It achieves parallel solution of multiple parameters such as equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient by relying on matrix synchronous operation, which greatly improves the real-time performance of parameter identification. It provides accurate and real-time cabinet thermal characteristic benchmark parameters for the quantitative evaluation of sealing performance during the door opening disturbance recovery period. The entire recursive calculation process does not require offline calibration without equipment shutdown, and is fully adapted to the usage requirements of uninterrupted storage operation of medical ultra-low temperature freezers.

[0046] like Figure 3 As shown, this process is specifically designed for identifying and calculating the thermal characteristic parameters of the cabinet under steady-state insulation conditions. The first unit is the input steady-state insulation time-series dataset. This unit receives the dedicated steady-state insulation time-series data from the upper-level time-segmentation unit. The data includes pre-processed equivalent average temperature time-series, ambient temperature time-series, compressor average power time-series, and compressor start / stop status marker sequences. All data has undergone Grubbs anomaly cleaning and filtering, eliminating abrupt noise data. Furthermore, all sampled frames meet the multiple judgment conditions for steady-state insulation conditions. Only continuous sampled data that meets the standards is sent to the identification process. Data from transient cooling and door opening disturbance recovery periods are not included in this process, thus avoiding interference from non-steady-state conditions in solving the thermal balance parameters. After the data input is completed, the process proceeds to constructing the thermal balance discrete difference regression vector. This unit utilizes the discrete form of the lumped-parameter heat balance equation to construct vectors. It extracts the derivative of temperature with respect to time from the equivalent average temperature difference between the current and previous sampling times. This is combined with the difference between the ambient temperature and the cabinet's equivalent average temperature to construct a heat transfer driving term. The compressor's average power is multiplied by the start / stop status marker to obtain the refrigeration input term. The three core variables are integrated into a standardized regression vector, with the internal variable dimensions corresponding one-to-one with the three types of thermal characteristic parameters to be solved, providing a standardized input matrix for subsequent iterative calculations. After the regression vector construction is complete, the prediction error and gain vector are calculated. The parameter estimation unit is divided into three layers of synchronous operation logic. First, it retrieves the predicted values ​​of thermal characteristic parameters after convergence in the previous sampling period, substitutes them into the heat balance equation to solve for the theoretical predicted value of the cabinet temperature at the current moment, calculates the difference between the theoretical predicted value and the actual equivalent average temperature collected by the system to obtain the prediction error, and then introduces a fixed forgetting factor and the parameter covariance matrix stored in the previous period to complete the numerical solution of the gain vector. The gain vector represents the correction weight of the current sampling data on the parameter prediction result. Finally, the gain vector and the prediction error are used to synchronously correct the predicted values ​​of the three sets of parameters: equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient, to complete the parameter update for a single sampling point.

[0047] After the parameters are updated, the process moves to updating the covariance matrix. The unit, the covariance matrix, records the uncertainty of the three sets of thermal characteristic parameter prediction results. It iterates synchronously with each parameter correction, and the matrix values ​​gradually converge and stabilize with each iteration, fully preserving the error fluctuation characteristics of each round of parameter identification. After the covariance matrix update, it enters the convergence judgment unit. This unit calculates the change magnitude of the three sets of parameter prediction values ​​relative to the previous round and compares the change magnitude with the system's preset convergence judgment threshold. If there is no questioning statement, and the parameter change magnitude exceeds the convergence threshold, it indicates that the parameter prediction results still have significant fluctuations and have not reached the stability standard. The process then flows to cache the current parameters and enters the next sampling cycle iteration unit. This unit temporarily stores the updated parameter values ​​and covariance matrix, and then returns to constructing the thermal equilibrium discrete difference regression vector through a loop. The unit reads the steady-state time-series data at the next sampling moment to start a new round of iterative calculation. If the parameter change amplitude remains below the convergence threshold for multiple rounds, it indicates that the three sets of thermal characteristic parameters have stabilized. The process then jumps to the unit that locks the equivalent thermal resistance / heat capacity / cooling efficiency coefficient. This unit stores the three sets of parameters after iterative convergence in the local non-volatile storage area. The stored parameters will not be cleared and re-identified until the next round of whole-machine testing cycle starts. After the parameters are solidified, they are transferred to the output thermal characteristic parameter signal to the parameter identification module unit. This unit encapsulates the stable equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient into standardized thermal characteristic parameter signals, which are simultaneously sent to the sealing performance index calculation unit and the whole-machine performance evaluation unit for the door opening disturbance recovery condition. The entire process relies on the sampling cycle to achieve online parameter identification. It distinguishes between continuous iteration and parameter solidification output based on the convergence judgment branch. Each calculation unit has a clear division of labor, which fully realizes the solution of thermal characteristic parameters without downtime, adapting to the operation requirements of uninterrupted storage operation of medical ultra-low temperature freezers.

[0048] The performance evaluation module, acting as the system's output aggregation layer, receives thermal characteristic parameter signals, sealing performance index signals, and preprocessed dataset signals. It then calculates temperature field quality indicators and determines the grade of all performance indicators. The module extracts effective sensor temperature data corresponding to the steady-state insulation period signal. It calculates the standard deviation of the temperature values ​​from all effective sensors at the same sampling moment as the instantaneous value of temperature uniformity at that moment. The module then arithmetically averages the instantaneous values ​​from all sampling moments within the steady-state insulation period to obtain the temperature uniformity signal, which reflects the degree of temperature difference at different locations within the cabinet at the same time. The module calculates the standard deviation of the time series of the equivalent average temperature within the steady-state insulation period to obtain the temperature fluctuation signal, reflecting the amplitude of temperature fluctuations within the cabinet around the target value over time. Finally, the module compares the equivalent thermal resistance signal, cooling efficiency coefficient signal, sealing performance index signal, temperature fluctuation signal, and temperature uniformity signal with their respective preset grading thresholds to determine the insulation performance grade, cooling efficiency grade, sealing performance grade, temperature stability grade, and temperature uniformity grade. The module then summarizes the grade results of these five dimensions into a comprehensive test report signal for output. This comprehensive test report transforms raw sensor data, previously scattered across various physical domains, into structured performance indicators for equipment health management. This allows users and maintenance personnel to readily grasp the current status and performance degradation trends of the equipment, providing a quantitative basis for developing preventative maintenance plans and early fault intervention. The entire testing process, from data acquisition to report generation, is automated under normal freezer operation conditions, requiring no sample emptying, no shutdown, and no additional specialized testing environment. This makes it exceptionally valuable for equipment like medical ultra-low temperature freezers that require continuous year-round operation. The formula for calculating the sealing performance index signal is as follows:

[0049] ;

[0050] Where S represents the real-time sealing performance index of the medical ultra-low temperature freezer. These are the weighting coefficients corresponding to the thermal resistance degradation factor. The weighting coefficients corresponding to the recovery time degradation factor are used. R is the factory-standard equivalent thermal resistance, and R is the real-time equivalent thermal resistance identified by the recursive least squares gain vector algorithm with forgetting factor in the second band. The door opening and recovery time is the factory standard. The measured door opening recovery time is the time taken during the period of door opening disturbance. This is the fundamental weighting constant for thermal resistance sensitivity. This is the thermal resistance aging degradation adjustment coefficient. This is an exponential correction term that varies with the degree of thermal resistance degradation. =1 is the weight normalization constraint. The factory baseline thermal resistance and factory baseline recovery time are pre-stored in the system's non-volatile memory. These are calibrated and recorded using the same recursive least squares algorithm under standard operating conditions for the same model of freezer. During the online testing phase, the ratio of the real-time equivalent thermal resistance to the factory baseline thermal resistance is defined as the thermal resistance degradation factor, and the ratio of the measured door opening recovery time to the factory baseline recovery time is defined as the recovery time degradation factor. The inverse of each degradation factor is used to amplify the exponential change caused by performance degradation, and then the weights of the two types of indicators are dynamically allocated through an exponential function. In the early stages of the freezer's service life, when the insulation layer has not shown significant aging, the thermal resistance degradation factor approaches 1. The weighting is relatively low, with the evaluation focusing primarily on the sealing effect of the door seal; as the equipment ages over long-term use and the equivalent thermal resistance decreases, the index item automatically increases. The weighting process takes into account both the effects of thermal insulation loss and cold leakage from the door seal. The weighting normalization constraint ensures that the output range of the index is stable and controllable. Finally, the weighted summation yields the sealing performance index, which is then sent to the performance evaluation module to complete the level determination. This system of equations fully incorporates all the core output data from the previous two algorithms. Grubbs' statistics ensure that the temperature sequence does not experience abnormal jumps during the door opening disturbance period, accurately extracting the temperature rise peak and door opening recovery time. The recursive least squares gain vector algorithm with forgetting factor provides high-confidence real-time equivalent thermal resistance. These three elements form a complete serial calculation link, solving the shortcomings of traditional fixed-weight linear weighted models that cannot adapt to the differences in sensitivity of indicators at different aging stages throughout the equipment's life cycle. A higher sealing performance index value indicates better overall thermal insulation performance of the freezer door seal and cabinet, which can be directly used to determine door seal aging and provide early warning of insulation layer failure. The output is used in the comprehensive test report as a core indicator for compliance verification of medical ultra-low temperature freezers. The entire algorithm link forms a closed-loop calculation from cleaning the original sensor data and identifying thermal characteristic parameters online to comprehensive sealing performance rating. All formulas are standardized and quantified to quantify equipment performance indicators, avoiding errors from subjective human judgment. This fully meets the industry regulatory requirements for standardized, traceable, and automated medical device testing, significantly reducing manual testing costs and improving the accuracy and efficiency of performance testing for medical low-temperature storage equipment.

[0051] This invention discloses a comprehensive performance testing system for medical ultra-low temperature freezers. It simultaneously collects multiple temperature readings, compressor power data, and door opening / closing signals within the freezer. After outlier removal and low-pass filtering, the system automatically divides the timeframe into steady-state insulation, transient cooling, and door opening disturbance periods based on door status and temperature change rate. During the steady-state period, a recursive least squares algorithm with a forgetting factor is used to solve the lumped-parameter thermal balance equation online, identifying the equivalent thermal resistance, heat capacity, and refrigeration efficiency coefficient. During the door opening disturbance period, the system extracts the temperature rise peak and recovery time, and calculates the sealing performance index based on thermal characteristic parameters. Simultaneously, the system calculates the standard deviation and maximum deviation of the effective sensor temperatures, generating temperature fluctuation and uniformity signals. Finally, a comprehensive test report is output, enabling quantitative assessment of performance degradation without interrupting freezer operation.

[0052] Therefore, the comprehensive performance testing system for medical ultra-low temperature freezers of the present invention can solve the problem that it is difficult to conduct online comprehensive testing of the insulation performance, door seal sealing performance and temperature uniformity of ultra-low temperature freezers during operation.

[0053] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A comprehensive performance testing system for medical ultra-low temperature freezers, characterized in that, include: The sensor acquisition module acquires synchronous timing data stream signals; The data preprocessing module receives synchronous time-series data stream signals, performs Grubbs criterion anomaly removal and interpolation on multi-point temperature signals, and performs low-pass filtering on compressor electrical parameter signals to form a preprocessed dataset signal. The state classification module receives the preprocessed dataset signal and, based on the door opening / closing state change edge and the comparison results of the temperature change rate with the threshold, divides the operating period into steady-state heat preservation, transient cooling, and door opening disturbance recovery period signals. The parameter identification module receives signals during steady-state insulation and door opening disturbance recovery periods. Under the steady-state insulation period signal, it solves the heat balance equation by recursive least squares to generate thermal characteristic parameter signals such as equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient. Under the door opening disturbance recovery period signal, it extracts the door closing temperature, peak rise, and recovery time, and combines the thermal characteristic parameters to generate a sealing performance index signal. The performance evaluation module receives thermal characteristic parameter signals, sealing performance index signals, and preprocessed dataset signals, calculates the effective temperature standard deviation to generate temperature fluctuation signals, and summarizes and outputs a comprehensive test report signal.

2. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 1, characterized in that, After receiving the synchronous time-series data stream signal, the data preprocessing module performs outlier detection based on the Grubbs criterion on each temperature data sequence in the multi-point temperature signal, marks the data points determined to be outliers and removes them from the sequence, and fills in the positions of the removed data points by linear interpolation of adjacent valid data points to form a continuous and complete temperature data sequence; the data preprocessing module performs low-pass filtering and smoothing on the average power signal in the compressor electrical parameter signal to remove high-frequency noise components, and combines the processed multi-point temperature data sequence with the smoothed compressor average power signal, door switch status signal and ambient temperature signal to form the preprocessed dataset signal.

3. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 2, characterized in that, Before performing outlier removal based on the Grubbs criterion on multi-point temperature signals, the data preprocessing module first calculates the arithmetic mean and standard deviation of all temperature sensor readings at the current moment. Using the standard deviation as a benchmark, it determines the critical coefficient corresponding to the number of sensors in the Grubbs criterion. The statistic obtained by dividing the absolute value of the deviation of each sensor reading from the arithmetic mean by the standard deviation is compared with the critical coefficient. Sensor readings exceeding the critical coefficient are judged as outliers and removed; sensor readings not exceeding the critical coefficient are retained as valid data. After completing outlier removal, the data preprocessing module renumbers and counts the retained valid sensors, using the number of valid sensors as input parameters for subsequent state classification and performance evaluation. When the number of valid sensors is lower than the preset minimum number of sensors, a sensor fault alarm signal is generated. The calculation process of comparing the statistic obtained by dividing the absolute value of the deviation of each sensor reading from the arithmetic mean by the standard deviation with the critical coefficient is as follows: ； in, Let be the Grubbs statistic corresponding to the i-th temperature sensor at the k-th sampling time. Here, N represents the raw temperature reading of the i-th sensor at the k-th sampling time, and N is the total number of temperature sensors connected at the current time. This is the summation of all sensor temperature readings at the current moment. This represents the sample standard deviation of all temperature readings at the current moment. The arithmetic mean of the current temperature. This represents the absolute deviation between the temperature of a single sensor and its mean value.

4. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 1, characterized in that, After receiving the preprocessed dataset signal, the state classification module monitors the logic level changes of the door switch status signal in real time. When it detects a change from the closed level to the open level, it records the opening time; when it detects a change from the open level to the closed level, it records the closing time and marks the continuous time period after the closing time as the door opening disturbance recovery period signal. Within the interval where the door switch status signal remains at the closed level, the state classification module performs frame-by-frame differential calculation on the equivalent average temperature obtained by fusing multi-point temperature signals after consistency verification, and calculates the temperature change rate per unit time. When the absolute value of the temperature change rate is greater than or equal to a preset cooling change rate threshold and the equivalent average temperature is higher than the set target temperature, the corresponding time period is marked as the transient cooling period signal; when the absolute value of the temperature change rate is less than a preset insulation change rate threshold and the equivalent average temperature fluctuates near the set target temperature, the corresponding time period is marked as the steady-state insulation period signal.

5. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 4, characterized in that, When classifying steady-state insulation period signals, the state classification module needs to meet the following conditions simultaneously: the door switch status signal remains at the closed level, the absolute value of the difference between the equivalent average temperature and the set target temperature is less than the preset temperature deviation allowable value, and the standard deviation of the equivalent average temperature in multiple consecutive sampling periods is less than the preset temperature fluctuation allowable value. The continuous time period in which all three conditions are met is marked as an effective steady-state insulation period signal. The state classification module outputs the start and end times of the effective steady-state heat preservation period signal to the parameter identification module, and identifies the start and stop status of the compressor electrical parameter signal during the effective steady-state heat preservation period. When the average power of the compressor is greater than the preset start power threshold, the compressor is determined to be in the running state. When the average power of the compressor is less than the preset standby power threshold, the compressor is determined to be in the stopped state. The compressor start and stop status sequence signal is generated as a component of the steady-state heat preservation period signal.

6. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 5, characterized in that, After receiving the steady-state heat preservation period signal, the parameter identification module extracts the equivalent average temperature sequence, ambient temperature sequence, and compressor average power sequence corresponding to the steady-state heat preservation period from the preprocessed dataset signal. The parameter identification module establishes a lumped parameter heat balance equation, approximates the derivative of the equivalent average temperature with respect to time by the ratio of the temperature difference between adjacent sampling times to the sampling interval, uses the difference between the ambient temperature and the equivalent average temperature as the heat transfer driving force term, and uses the product of the compressor average power and the compressor start-stop state sequence signal as the refrigeration input term, thus forming a discrete-time difference equation containing three unknown parameters: equivalent thermal resistance, equivalent heat capacity, and refrigeration efficiency coefficient. The parameter identification module uses a recursive least squares algorithm with a forgetting factor to perform online parameter estimation for the discrete-time difference equation. The value of the forgetting factor gives the algorithm greater weight to recent data. The parameter estimates are updated recursively by sampling point. When the relative change of the parameter estimates is less than the preset convergence threshold for several consecutive sampling periods, the converged equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient are output as the thermal characteristic parameter signals.

7. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 6, characterized in that, The parameter identification module employs a recursive least squares algorithm with a forgetting factor, which includes the following recursive steps: At each sampling time, a regression vector is constructed based on the current equivalent average temperature, the previous equivalent average temperature, the ambient temperature, the compressor average power, and the compressor start / stop status. The predicted temperature value at the current time is calculated using the parameter estimate value at the previous time. The difference between the actual equivalent average temperature at the current time and the predicted temperature value is used as the prediction error. The gain vector at the current time is calculated by combining the forgetting factor and the covariance matrix at the previous time. The parameter estimate value is corrected and updated using the gain vector and the prediction error, and the covariance matrix is ​​updated simultaneously. When the change in the parameter estimate is lower than the preset convergence threshold for multiple consecutive moments during the recursive process, the current parameter estimate is locked as the final equivalent thermal resistance, equivalent heat capacity, and cooling efficiency coefficient, and stored permanently until the parameter identification is triggered again in the next detection cycle. The calculation formula for the gain vector is as follows: ； in, This is the recursive least squares gain vector with forgetting factor at the k-th sampling time. Let be the parameter covariance matrix at the (k-1)th sampling time. The thermal equilibrium regression vector is constructed based on the filtered effective temperature sequence at the k-th sampling time. The preset forgetting factor value is 0 < <1, Let be the transpose of the regression vector. It is the scalar term of the product of the regression vector, covariance matrix, and transpose vector.

8. The comprehensive performance testing system for a medical ultra-low temperature freezer according to claim 4, characterized in that, After receiving the door opening disturbance recovery period signal, the parameter identification module extracts the equivalent average temperature at the door closing moment, the equivalent average temperature sequence during the door opening disturbance recovery period, and the compressor average power sequence from the preprocessed dataset signal. The parameter identification module traverses the equivalent average temperature sequence during the door opening disturbance recovery period and finds the temperature maximum value during the temperature recovery process as the temperature recovery peak value. The parameter identification module calculates the difference between the temperature recovery peak value and the set target temperature as the temperature recovery amplitude, and calculates the time required for the equivalent average temperature to fall back to the sum of the set target temperature and the preset recovery deviation value, starting from the time corresponding to the temperature recovery peak value, as the door opening recovery time. The parameter identification module calculates the theoretical thermal time constant as the product of the equivalent thermal resistance and equivalent heat capacity based on the identified equivalent thermal resistance and equivalent heat capacity, and uses the ratio of the theoretical thermal time constant to the door opening recovery time as the reference recovery rate. The reference recovery rate and the inverse change rate of the equivalent thermal resistance relative to the factory reference thermal resistance are weighted and combined to generate the sealing performance index signal.

9. A comprehensive performance testing system for medical ultra-low temperature freezers according to claim 8, characterized in that, When generating the sealing performance index signal, the parameter identification module incorporates the factory reference thermal resistance and factory reference recovery time as reference standards. These factory reference thermal resistance and recovery time are pre-stored in the system's non-volatile memory and are identified and recorded by the parameter identification module during factory testing of the same model of refrigerator under standard operating conditions using the same recursive least squares algorithm as online testing. The parameter identification module uses the ratio of the equivalent thermal resistance obtained online to the factory reference thermal resistance as a thermal resistance degradation factor, and the ratio of the door opening recovery time measured online to the factory reference recovery time as a recovery time degradation factor. The reciprocals of the thermal resistance degradation factor and the recovery time degradation factor are weighted and summed, with the weights allocated according to the sensitivity differences of the thermal resistance degradation factor and the recovery time degradation factor at different aging stages of the refrigerator. The weighted sum is the sealing performance index signal, calculated using the following formula: ； Where S represents the real-time sealing performance index of the medical ultra-low temperature freezer. These are the weighting coefficients corresponding to the thermal resistance degradation factor. The weighting coefficients corresponding to the recovery time degradation factor are used. R is the factory-standard equivalent thermal resistance, and R is the real-time equivalent thermal resistance identified by the recursive least squares gain vector algorithm with forgetting factor in the second band. The door opening and recovery time is the factory standard. The measured door opening recovery time is the time taken during the period of door opening disturbance. This is the fundamental weighting constant for thermal resistance sensitivity. This is the thermal resistance aging degradation adjustment coefficient. This is an exponential correction term that varies with the degree of thermal resistance degradation. =1 is the weight normalization constraint.

10. A comprehensive performance testing system for medical ultra-low temperature freezers according to claim 1, characterized in that, The performance evaluation module receives the preprocessed dataset signal, extracts the effective sensor temperature data corresponding to the steady-state heat preservation period signal, calculates the standard deviation of the temperature values ​​of all effective sensors at the same sampling time, uses the standard deviation as the instantaneous value of temperature uniformity at that sampling time, and performs an arithmetic mean of the instantaneous values ​​of temperature uniformity at all sampling times within the steady-state heat preservation period to generate the temperature uniformity signal; the performance evaluation module calculates the standard deviation of the time series of equivalent average temperature within the steady-state heat preservation period to generate the temperature fluctuation signal. The performance evaluation module compares the equivalent thermal resistance signal with a preset insulation performance grading threshold to determine the insulation performance level, compares the cooling efficiency coefficient signal with a preset cooling efficiency grading threshold to determine the cooling efficiency level, compares the sealing performance index signal with a preset sealing performance grading threshold to determine the sealing performance level, compares the temperature fluctuation signal with a preset fluctuation grading threshold to determine the temperature stability level, and compares the temperature uniformity signal with a preset uniformity grading threshold to determine the temperature uniformity level. The insulation performance level, cooling efficiency level, sealing performance level, temperature stability level, and temperature uniformity level are then combined to form the comprehensive test report signal.

Images