Blood filter equipment early warning method based on multi-parameter coupling and blood filter

Abstract

The invention discloses a blood filter early warning method based on multi-parameter coupling and a blood filter, and aims to solve the problems of lagging and high false alarm rate of the existing single parameter threshold alarm. The method comprises the following steps: acquiring time sequence data of at least two monitoring parameters such as transmembrane pressure and venous pressure in real time; constructing a risk state vector based on the sequential data; calculating a risk covariance matrix of the vector in a set time window so as to quantify a collaborative relationship of variation trends among parameters; performing matching degree comparison on the matrix and a pre-stored standard fault feature matrix; and generating and outputting early warning information at least indicating one equipment abnormal mode based on the comparison result. By analyzing a multi-parameter dynamic cooperative relationship instead of a single parameter absolute value, early warning and intelligent preliminary diagnosis of filter blood coagulation, pipeline abnormity and other faults are realized, and the timeliness, accuracy and clinical decision support capability of alarm are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of blood filtration machine technology, and in particular to a blood filtration equipment early warning method and blood filtration machine based on multi-parameter coupling. Background Technology

[0002] Continuous renal replacement therapy (CRRT) is a key technology for the treatment of critically ill patients, and hemofiltration machines are its core equipment. Currently, hemofiltration machines generally use alarm mechanisms based on a single static threshold parameter, which has significant drawbacks.

[0003] First, this mechanism suffers from a significant time lag. For example, filter clotting is a slow process, with the core indicator, transmembrane pressure (TMP), potentially rising from 80 mmHg to 150 mmHg over several tens of minutes. Currently, the alarm threshold is typically set at 180 mmHg, meaning an alarm is only triggered when clotting has become quite severe, missing the optimal window for preventative intervention. In other words, for existing hemofiltration machines, the alarm is only triggered when parameters exceed a fixed threshold. By this time, the malfunction (such as filter clotting) has often progressed to a middle or late stage, leaving healthcare professionals with a very short window for effective intervention, forcing them to take reactive measures and easily leading to treatment interruptions.

[0004] Secondly, it suffers from a high false alarm rate. For example, when a patient coughs or turns over, causing a brief fluctuation in central venous pressure, it can directly trigger an alarm on the hemofiltration machine due to a momentary spike in venous pressure, even though this is not a malfunction of the equipment or treatment. The brief fluctuation in venous pressure caused by a patient's cough triggers an invalid alarm, while genuine early signs of risk are drowned out. This leads to "alarm fatigue," reducing the alertness of healthcare workers to alarms. In other words, the system judges parameters in isolation, ignoring their inherent correlations, resulting in a large number of interfering "false alarms."

[0005] Finally, existing alarms lack intelligent analysis. They can only indicate "where the abnormality is," but cannot determine "why the abnormality is," and root cause investigation relies entirely on personal experience, which can easily delay handling in emergency situations.

[0006] Therefore, it is necessary to improve the existing alarm method of blood filtration machines to solve the problems of serious delay in early warning, high false alarm rate and inability to provide intelligent decision support caused by the use of a single parameter static threshold alarm.

[0007] The information disclosed in this background section is included only to enhance the understanding of the context of this disclosure, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] One objective of this invention is to provide a blood filtration equipment early warning method and blood filtration machine based on multi-parameter coupling, which can solve the problems of serious early warning delay, high false alarm rate and inability to provide intelligent decision support caused by the use of single-parameter static threshold alarm in existing blood filtration machine alarm methods.

[0009] To achieve the above objectives, in one aspect, the present invention provides a pre-warning method for blood filtration equipment based on multi-parameter coupling, comprising: S1. Real-time acquisition of time-series data of at least two risk status indicators of the blood filtration machine operation, wherein the at least two risk status indicators include transmembrane pressure and venous pressure; S2. Construct a risk state vector based on the time-series data of the at least two risk state indicators; S3. Based on the changes of the risk state vector within a set time window, calculate a risk covariance matrix, wherein the risk covariance matrix is ​​used to characterize the synergistic relationship between the changing trends of the at least two risk state indicators. S4. Compare the calculated risk covariance matrix with at least one pre-stored standard fault feature matrix to determine the matching degree, wherein different standard fault feature matrices are associated with different equipment anomaly modes. S5. Based on the matching degree comparison result, generate and output early warning suggestion information; wherein, the early warning suggestion information indicates at least one abnormal device mode and corresponding operation suggestions.

[0010] Optionally, in step S2, the constructed risk state vector R(t) is represented as:

[0011] in, n: a positive integer not less than 2; [x1(t), x2(t), ..., x n [(t)]: The values ​​of n risk status indicators determined based on the time series data of the at least two risk status indicators at time t; [x1(t), x2(t), ..., x n (t)] T : represents [x1(t), x2(t), ..., x n The transpose of [(t)].

[0012] Optional, The n risk status indicators include at least two of the following: the instantaneous value of the transmembrane pressure, the rate of change of the transmembrane pressure, the instantaneous value of the venous pressure, the rate of change of the venous pressure, a risk integral function value constructed based on the transmembrane pressure, a risk integral function value constructed based on the venous pressure and arterial pressure, and the rate of change of the pressure difference between the venous pressure and the arterial pressure.

[0013] Optionally, the risk integral function F1(t) constructed based on the transmembrane pressure is defined as:

[0014] in, T is the start time of integration, and t is the end time of integration; Transmembrane pressure as a function of time A changing function; for Rate of change per unit time; This is a mapping function used to convert the rate of change of transmembrane pressure. This is mapped to a risk contribution rate for accumulation.

[0015] Optionally, a risk integral function is constructed based on the venous pressure and arterial pressure. Defined as:

[0016] in, T is the start time of integration, and t is the end time of integration; This refers to the pressure difference between the venous and arterial ends of the extracorporeal circulation loop.

[0017] Blood flow velocity; This is a mapping function used to represent the rate of change of flow resistance. This is mapped to a risk contribution rate for accumulation.

[0018] Optionally, step S3 includes

[0019] Define a continuous historical time period, i.e., a time window [tW, t], where t is the current time and W is the preset window length; Obtain a series of values ​​for the risk state vector R(τ) within the time window [tW, t], a total of N observation samples arranged in chronological order, denoted as R(1), R(2), …, R(N); where each observation sample R(k) (k=1,2,...,N) is an n-dimensional column vector. This represents the value of the i-th risk status index at the k-th sampling time; Construct an n × n risk covariance matrix C(t) based on these N samples; The steps for constructing the risk covariance matrix C(t) include: First, calculate the sample mean of each risk status indicator within the time window. ;

[0020] Then, each element in the risk covariance matrix C(t) is calculated. ;in, The covariance between the i-th and j-th risk status indicators within the time window is represented by the following formula: Where; when i = j, the calculated This is the sample variance of the i-th risk status indicator within that time window: ; All calculated elements This forms an n × n real symmetric matrix, namely the risk covariance matrix C(t).

[0021] Optionally, step S4 specifically includes: S41. Calculate the risk covariance matrix respectively. Matrix distance between each of the pre-stored standard fault feature matrices ,in The total number of standard fault feature matrices; wherein different standard fault feature matrices are associated with different equipment anomaly modes; S42. Determine the minimum matrix distance. and its corresponding standard fault feature matrix ; S43. Determine the minimum matrix distance. Is it less than the preset matching threshold?

[0022] Optionally, in step S41,

[0023] in, Real-time risk covariance matrix The element in row p and column q; Standard Fault Feature Matrix The element in row p and column q; n: Matrix dimension, equal to the number of risk status indicators.

[0024] Optionally, step S5 includes: S51. Determine whether the minimum matrix distance is less than a preset matching threshold; S52. When the determination is yes, the device anomaly mode associated with the standard fault feature matrix corresponding to the minimum matrix distance is determined as the currently identified target fault mode. S53. Generate and output early warning suggestion information, wherein the early warning suggestion information includes a textual description of the target fault mode and operation suggestions corresponding to the target fault mode.

[0025] On the other hand, a blood filtration machine is provided for executing any of the aforementioned blood filtration device early warning methods based on multi-parameter coupling, comprising: The data monitoring module is used to acquire time-series data of at least two risk status indicators of the blood filtration machine in real time, wherein the at least two risk status indicators include transmembrane pressure and venous pressure; A vector construction module is used to construct a risk state vector based on the time-series data of the at least two risk state indicators. The matrix calculation module is used to calculate a risk covariance matrix based on the changes of the risk state vector within a set time window, wherein the risk covariance matrix is ​​used to characterize the synergistic relationship between the changing trends of the at least two risk state indicators. The matching module is used to compare the calculated risk covariance matrix with at least one pre-stored standard fault feature matrix, wherein different standard fault feature matrices are associated with different equipment anomaly modes. The early warning suggestion module is used to generate and output early warning suggestion information based on the matching degree comparison results; wherein, the early warning suggestion information indicates at least one abnormal device mode and corresponding operation suggestions.

[0026] The beneficial effects of this invention are as follows: It provides a pre-warning method for blood filtration equipment based on multi-parameter coupling, and the core working process is as follows: First, in step S1, the system continuously acquires time-series data of at least two risk status indicators generated during the operation of the hemofiltration machine. These risk status indicators are key physical quantities reflecting the status of the extracorporeal circulation loop, which explicitly include transmembrane pressure, which is extremely sensitive to the filter status, and venous pressure, which is sensitive to loop patency. Then, in step S2, the system dynamically constructs a risk state vector based on these real-time acquired time-series data. This risk state vector is a mathematical set that organizes multiple risk state indicators or their derived indicators (such as rate of change, calculated values, etc.) at a given moment into an ordered array, thereby mathematically representing the comprehensive state of the system at that moment. Next, in step S3, the core innovation of the method is demonstrated: the system does not analyze individual elements in the vector in isolation, but calculates a risk covariance matrix based on the historical data sequence of the risk state vector within a set time window. This risk covariance matrix is ​​a square matrix, and each element quantifies the cooperative relationship of the changing trends of different risk state indicators within the time window, such as the correlation strength between the rising trend of transmembrane pressure and the rising trend of venous pressure; Subsequently, in step S4, the system compares the risk covariance matrix, which is calculated in real time and represents the current "cooperative change pattern," with at least one standard fault feature matrix in a pre-stored database. Each standard fault feature matrix is ​​like a "fingerprint," pre-learned or defined from historical data of a specific, known equipment anomaly pattern (such as simple filter clotting, excessively high venous reservoir fluid level, or poor arterial drainage). By comparison, the standard fault most similar to the current pattern can be identified.

[0027] Finally, in step S5, the system generates and outputs early warning suggestion information based on the matching degree comparison results. This early warning suggestion information differs from traditional alarms that only inform that "a certain parameter exceeds the limit," but it can indicate at least one of the most likely abnormal device modes, such as "early filter clotting mode" or "early venous circuit obstruction mode," etc.

[0028] The multi-parameter coupling-based early warning method for blood filtration equipment provided in this embodiment fundamentally changes the logical basis of traditional alarms. Traditional methods rely on the instantaneous absolute value of a single parameter, while this invention, by constructing and analyzing a risk covariance matrix, shifts to a deeper exploration of the dynamic synergistic relationships between multiple parameters. This enables the system to keenly detect early patterns of "synergistic deterioration" among multiple parameters before the instantaneous value of any risk status indicator reaches its static alarm threshold, thus achieving true early warning. Simultaneously, by matching real-time patterns with a standard fault feature matrix library, a leap from "real-time alarm" to "preliminary early warning" is achieved, providing medical personnel with more operational fault cause identification and effectively solving the problems of traditional alarm delays, high false alarm rates, and lack of intelligent analysis.

[0029] Therefore, the blood filtration equipment early warning method and blood filtration machine based on multi-parameter coupling provided by the present invention can solve the problems of serious early warning delay, high false alarm rate and inability to provide intelligent decision support caused by the use of single-parameter static threshold alarm in existing blood filtration machine alarm methods. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0031] Figure 1 A flowchart of a blood filtration device early warning method based on multi-parameter coupling provided for an embodiment; Figure 2 The structural block diagram of the blood filtration machine provided in the embodiment. Detailed Implementation

[0032] In this invention, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment can be included in at least one embodiment of the invention. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this invention, as long as there is no technical contradiction or conflict, the technical features mentioned in each embodiment can be combined in any way to form a corresponding implementable technical solution.

[0033] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit the invention.

[0034] In the description of this invention, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " generally indicates that the preceding and following objects have an "or" logical relationship.

[0035] In this invention, terms such as “first” and “second” are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy, or order between these entities or operations.

[0036] Without further limitations, the use of terms such as “comprising,” “including,” “having,” or other similar expressions in this invention is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0037] Similar to the understanding in the Examination Guidelines, in this invention, expressions such as "greater than," "less than," and "exceeding" are understood to exclude the stated number; expressions such as "above," "below," and "within" are understood to include the stated number. Furthermore, in the description of the embodiments of this invention, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times," unless otherwise explicitly specified.

[0038] In the description of the embodiments of the present invention, the spatial related expressions used, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "vertical," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," "circumferential," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or drawings. They are only for the purpose of describing the specific embodiments of the present invention or for the reader's understanding, and do not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0039] Unless otherwise explicitly stated or limited, the terms "installation," "connection," "linking," "fixing," and "setting," as used in the description of the embodiments of this invention, should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral arrangement; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this invention according to the specific circumstances.

[0040] This invention provides a blood filtration equipment early warning method and blood filtration machine based on multi-parameter coupling, which can solve the problems of serious early warning delay, high false alarm rate and inability to provide intelligent decision support caused by the use of single-parameter static threshold alarm in existing blood filtration machine alarm methods.

[0041] Example 1 See Figure 1 This embodiment provides a pre-warning method for blood filtration equipment based on multi-parameter coupling, including: S1. Real-time acquisition of time-series data of at least two risk status indicators of the blood filtration machine operation, wherein the at least two risk status indicators include transmembrane pressure and venous pressure; S2. Construct a risk state vector based on the time-series data of the at least two risk state indicators; S3. Based on the changes of the risk state vector within a set time window, calculate a risk covariance matrix, wherein the risk covariance matrix is ​​used to characterize the synergistic relationship between the changing trends of the at least two risk state indicators. S4. Compare the calculated risk covariance matrix with at least one pre-stored standard fault feature matrix to determine the matching degree, wherein different standard fault feature matrices are associated with different equipment anomaly modes. S5. Based on the matching degree comparison result, generate and output early warning suggestion information; wherein, the early warning suggestion information indicates at least one abnormal device mode and corresponding operation suggestions.

[0042] The core working process of the blood filtration machine early warning method based on multi-parameter coupling provided in this embodiment is as follows: First, in step S1, the system continuously acquires time-series data of at least two risk status indicators generated during the operation of the hemofiltration machine. These risk status indicators are key physical quantities reflecting the status of the extracorporeal circulation loop, which explicitly include transmembrane pressure, which is extremely sensitive to the filter status, and venous pressure, which is sensitive to loop patency. Then, in step S2, the system dynamically constructs a risk state vector based on these real-time acquired time-series data. This risk state vector is a mathematical set that organizes multiple risk state indicators or their derived indicators (such as rate of change, calculated values, etc.) at a given moment into an ordered array, thereby mathematically representing the comprehensive state of the system at that moment. Next, in step S3, the core innovation of the method is demonstrated: the system does not analyze individual elements in the vector in isolation, but calculates a risk covariance matrix based on the historical data sequence of the risk state vector within a set time window. This risk covariance matrix is ​​a square matrix, and each element quantifies the cooperative relationship of the changing trends of different risk state indicators within the time window, such as the correlation strength between the rising trend of transmembrane pressure and the rising trend of venous pressure; Subsequently, in step S4, the system compares the risk covariance matrix, which is calculated in real time and represents the current "cooperative change pattern," with at least one standard fault feature matrix in a pre-stored database. Each standard fault feature matrix is ​​like a "fingerprint," pre-learned or defined from historical data of a specific, known equipment anomaly pattern (such as simple filter clotting, excessively high venous reservoir fluid level, or poor arterial drainage). By comparison, the standard fault most similar to the current pattern can be identified.

[0043] Finally, in step S5, the system generates and outputs early warning suggestion information based on the matching degree comparison results. This early warning suggestion information differs from traditional alarms that only inform that "a certain parameter exceeds the limit," but it can indicate at least one of the most likely abnormal device modes, such as "early filter clotting mode" or "early venous circuit obstruction mode," etc.

[0044] The multi-parameter coupling-based early warning method for blood filtration equipment provided in this embodiment fundamentally changes the logical basis of traditional alarms. Traditional methods rely on the instantaneous absolute value of a single parameter, while this invention, by constructing and analyzing a risk covariance matrix, shifts to a deeper exploration of the dynamic synergistic relationships between multiple parameters. This enables the system to keenly detect early patterns of "synergistic deterioration" among multiple parameters before the instantaneous value of any risk status indicator reaches its static alarm threshold, thus achieving true early warning. Simultaneously, by matching real-time patterns with a standard fault feature matrix library, a leap from "real-time alarm" to "preliminary early warning" is achieved, providing medical personnel with more operational fault cause identification and effectively solving the problems of traditional alarm delays, high false alarm rates, and lack of intelligent analysis.

[0045] In this embodiment, the risk state vector R(t) constructed in step S2 is represented as:

[0046] in, n: a positive integer not less than 2; [x1(t), x2(t), ..., x n [(t)]: The values ​​of n risk status indicators determined based on the time series data of the at least two risk status indicators at time t; [x1(t), x2(t), ..., x n (t)] T : represents [x1(t), x2(t), ..., x n The transpose of [(t)] indicates that the vector is typically used in subsequent mathematical operations as a column vector.

[0047] n is a positive integer not less than 2, representing the number of selected risk status indicators, reflecting the scalability of risk assessment dimensions. x1(t), x2(t), …, x n (t) represents the specific values ​​of the 1st, 2nd, and up to the nth risk status indicators at time t. These values ​​are determined based on the original time-series data of at least two of the aforementioned risk status indicators, through direct reading, instantaneous calculation, or historical derivation.

[0048] During the work process, a risk state vector is formed. The n risk status indicators are selected from a predefined set of physical or computational quantities closely related to the key risks of the blood filtration machine. These indicators must include at least two of the following categories: One category consists of instantaneous physical quantities that are directly measured or simply derived, such as the instantaneous value of transmembrane pressure, the rate of change of transmembrane pressure (which can be obtained through numerical differentiation), the instantaneous value of venous pressure, the rate of change of venous pressure, and the rate of change of the pressure difference between venous and arterial pressure. Another type is a comprehensive indicator that reflects the risk accumulation process, obtained through calculations such as integration. Examples include risk integral function values ​​based on transmembrane pressure and risk integral function values ​​based on venous pressure and arterial pressure.

[0049] When constructing the risk state vector, the system selects at least two different types of indicators from the aforementioned set of indicators and combines them. For example, a vector can simultaneously include the "transmembrane pressure change rate" reflecting the filter state, the "arteriovenous pressure difference change rate" reflecting the tubing state, and a comprehensive "coagulation risk integral function value". This combination ensures that the vector can characterize the system's risk state from different perspectives (instantaneous and cumulative, direct and indirect).

[0050] By specifying the composition of risk state indicators, the comprehensiveness and effectiveness of the information contained in the risk state vector are ensured. Selecting dynamic trend indicators such as "rate of change" and "risk integral function value," rather than merely instantaneous parameter values, allows the vector to carry "trend" information about the system's state evolution, which is crucial for early warning. This combined indicator selection strategy enables the subsequently calculated risk covariance matrix to more accurately capture the inherent coupling relationships between different risk dimensions, thereby improving the accuracy and reliability of failure mode identification.

[0051] Furthermore, the risk integral function F1(t) constructed based on the transmembrane pressure is defined as:

[0052] in, T is the integration start time, which can usually be the start time of the current blood filtration process performed by the blood filtration machine or a certain reset time; Transmembrane pressure as a function of time A changing function; for The rate of change per unit time directly reflects how fast the pressure rises forward of the filter and is the direct kinetic expectation of the coagulation process. This is a mapping function that maps the rate of change of transmembrane pressure. This is mapped to a corresponding risk contribution rate for accumulation.

[0053] For example, It can be a linear function, or a positive function that is only sensitive to the positive rate of change. (Integral symbol) This represents the continuous accumulation (integration) of the mapped risk contribution rate from the initial time T to the current time t. Therefore, the function value F1(t) physically represents the total coagulation risk accumulated due to the upward trend of transmembrane pressure during the time interval from time T to t. It is a scalar; the larger the value, the higher the historically accumulated coagulation risk.

[0054] The risk integral function F1(t) cleverly transforms the instantaneous trend of transmembrane pressure into a cumulative historical risk through integration. Compared to simply observing instantaneous values ​​or comparing thresholds, this can detect a continuous but slow deterioration process much earlier. For example, even if the current absolute value of transmembrane pressure is not high, as long as it maintains a low positive rate of change for a period of time, its risk integral value will steadily increase and trigger an alarm when it reaches the warning threshold. This is equivalent to providing the system with "memory" and "summarization" capabilities, enabling it to identify those gradual risks that "boil the frog in slowly boiling water," greatly improving the timeliness of warnings.

[0055] Accordingly, the risk integral function F2(t) constructed based on the venous pressure and arterial pressure is defined as:

[0056] in, T is the start time of integration, and t is the end time of integration; This refers to the pressure difference between the venous and arterial ends of the extracorporeal circulation loop; furthermore, For venous pressure, The pressure difference (PV-PA) represents arterial pressure. According to basic principles of fluid mechanics, under relatively stable blood flow velocity, the pressure difference (PV-PA) primarily corresponds to the resistance overcome by blood flowing through the entire extracorporeal circulation circuit (including filters). Therefore, the ratio (PV-PA) / QB can be approximated as a reflection of the total flow resistance of the entire extracorporeal circulation tubing. This involves calculating the rate of change of flow resistance (PV-PA) / QB over time. Blood flow velocity; This is a mapping function used to represent the rate of change of flow resistance. This is mapped to a corresponding risk contribution rate for accumulation. Integration The risk contribution rate is accumulated from time T to t to obtain the risk integral value F2(t). This function value F2(t) quantifies the total risk accumulated due to the increased flow resistance of the extracorporeal circulation tubing (which may be caused by various reasons such as filter clotting, tubing kinking, and venous chamber clotting).

[0057] The risk integral function F2(t) creatively introduces a composite physiological parameter—(PV-PA) / QB—as an approximate estimate of flow resistance and integrates its rate of change. This method is more robust than monitoring venous pressure or arteriovenous pressure differential alone. For example, when a physician adjusts the blood flow velocity QB, venous pressure and pressure differential will change even if the tubing condition remains unchanged. However, using the ratio (PV-PA) / QB can offset the effect of blood flow velocity changes to some extent, more purely reflecting changes in tubing patency. Integrating its rate of change allows for continuous tracking of the deterioration trend of tubing flow resistance. Whether it's a gradual increase due to slow clotting or a sudden, sharp change due to a sudden kink, both can be effectively captured and accumulated in the risk integral, thus achieving earlier and more accurate assessment of tubing-related risks. This is a significant improvement and supplement to alarm logic based on a single pressure parameter.

[0058] Furthermore, it should be noted that the mapping function and Its function is to map the rate of change (or other derived quantities) of risk status indicators into a "risk contribution value" used for risk scoring.

[0059] by For example, the following are several possible implementations of the mapping function: 1. Linear mapping function in, If it is a positive constant, then the risk integral is the linear accumulation of the rate of change. 2. Linear mapping function with threshold processing in, It is a positive threshold; only when the rate of change x exceeds the threshold. Risks accumulate only in time; 3. Nonlinear mapping function Where p>1 represents the exponential coefficient; this can amplify the contribution of larger rates of change and is more sensitive to rapid changes. 4. Piecewise linear mapping function

[0060] in, , For threshold ( < ), , For different intervals, the slope (usually) > ); 5. Mapping function trained based on historical data The specific form of the mapping function f1(x) can be obtained by training on historical monitoring data using machine learning methods (such as neural networks, support vector machines, etc.). For example, f1(x) can be a neural network whose input is the rate of change x and whose output is the risk contribution value. The structure and parameters of the neural network are determined through training on historical data to optimize early warning performance.

[0061] In practical implementation, the mapping function can be obtained through the following steps: ① Data collection: Collect a large amount of historical blood filtration machine operation data, including time-series monitoring data under normal conditions and various fault conditions (such as coagulation, tubing abnormalities, etc.); ② Feature extraction: Calculate the rate of change for each risk status indicator (e.g., obtain d(TMP) / dτ through difference or numerical differentiation). ③ Label definition: Based on the time point of the failure, label the data at each time point with the risk level (e.g., 0: no risk, 1: low risk, 2: high risk). ④ Function form selection: Based on clinical experience and data analysis, select a form of mapping function (such as linear, nonlinear, piecewise function, etc.). ⑤ Parameter Training: Optimize the parameters of the mapping function to minimize the difference between the predicted risk and the labeled risk. This can be achieved using supervised learning algorithms. ⑥ Validation and Testing: Use an independent dataset to validate the warning performance of the trained mapping function, and adjust the function form or parameters based on the test results.

[0062] In this embodiment, step S3 includes: S31. Determine a continuous historical time period, i.e., a time window [tW, t], where t is the current time and W is the preset window length; S32. Obtain a series of values ​​for the risk state vector R(τ) within the time window [tW, t], a total of N observation samples arranged in chronological order, denoted as R(1), R(2), …, R(N); where each observation sample R(k) (k=1,2,...,N) is an n-dimensional column vector. , This represents the value of the i-th risk status index at the k-th sampling time; S33. Construct an n × n risk covariance matrix C(t) based on these N samples; The steps for constructing the risk covariance matrix C(t) include: S331. First, calculate the sample mean of each risk status indicator within the time window. ;

[0063] S332. Then, calculate each element in the risk covariance matrix C(t). ;in, This characterizes the covariant relationship between the i-th risk state indicator and the j-th risk state indicator within the time window. The calculation formula is as follows: ;

[0064] Specifically, when i = j, the calculated This is the sample variance of the i-th risk status indicator within that time window: ;

[0065] S333, All calculated elements This forms an n × n real symmetric matrix, namely the risk covariance matrix C(t).

[0066] For the elements in the risk covariance matrix C(t): ① The diagonal element c in the matrix ii : represents the "volatility" (i.e., variance) of the i-th risk indicator itself, such as c 11 It is the magnitude of the fluctuation of the instantaneous transmembrane pressure within the time window; ② The off-diagonal elements c in the matrix ij : Represents the degree of "covariance" (i.e., the degree of covariance) between the i-th and j-th risk indicators: like The two indicators rise and fall together (for example, when transmembrane pressure increases, venous pressure also increases). like Two indicators "rise while fall" (for example, when transmembrane pressure increases, blood flow velocity decreases); like The changes in the two indicators are not correlated (e.g., there is no significant correlation between changes in transmembrane pressure and changes in arterial pressure). It should be noted that since time t is constantly updated dynamically, each element of C(t) is also updated in real time and on a rolling basis. As a new time t arrives, the window slides forward, old data is removed, new data is included, and the risk covariance matrix is ​​recalculated, thereby realizing dynamic and continuous monitoring of the collaborative relationship between multiple parameters.

[0067] When there are multiple risk status indicators (such as transmembrane pressure, venous pressure, arterial pressure, etc.), the trend of a single parameter (such as an increase in transmembrane pressure) may not be sufficient to determine the malfunction (for example, filter coagulation may be accompanied by both an increase in transmembrane pressure and fluctuations in venous pressure). The core role of the risk covariance matrix is ​​to quantify the "cooperative relationship" between multiple parameters - such as "whether venous pressure also increases synchronously when transmembrane pressure increases" and "how strong the correlation of the magnitude of change is". These relationships cannot be reflected by time series data of a single parameter alone.

[0068] By explicitly employing time-window-based sample covariance calculation, a standard, reliable, and implementable mathematical tool is provided for the abstract concept of "quantifying the synergistic relationship between parameters." The time-window mechanism allows the system to focus on recent dynamic changes, fully utilizing historical trend information while avoiding interference from outdated data in current state assessments. The risk covariance matrix C(t) generated by this dynamic calculation acts like a real-time updated "relationship graph," keenly reflecting the latest state of internal coupling relationships within the system. This provides accurate, time-varying input features for subsequent fault mode matching, enabling the entire early warning method to achieve dynamic early warning.

[0069] In this embodiment, step S4 specifically includes: S41. Calculate the risk covariance matrix respectively. Matrix distance between each of the pre-stored standard fault feature matrices ,in, M is the total number of standard fault feature matrices; where different standard fault feature matrices are associated with different equipment anomaly modes. Furthermore,

[0070] in, Real-time risk covariance matrix The element in row p and column q; Standard Fault Feature Matrix The element in row p and column q; n: Matrix dimension (equal to the number of risk status indicators); Used for calculation The physical model described above, with its specific numerical values, comprehensively considers the differences in every element (i.e., every pair of parameter synergies) within the risk covariance matrix, providing a global similarity measure. This is more comprehensive in capturing the overall differences between two "synergy patterns" than methods that only compare partial eigenvalues ​​or diagonal elements.

[0071] S42. Determine the minimum matrix distance. and its corresponding standard fault feature matrix ; S43. Determine the minimum matrix distance. Is it less than the preset matching threshold?

[0072] In substep S41, the system traverses all M pre-stored standard fault feature matrices. For each one, calculate the distance between the current real-time risk covariance matrix C(t) and its matrix. This distance value It is a scalar that quantifies the degree of difference between the currently observed cooperative mode and the i-th standard failure mode.

[0073] Next, in sub-step S42, the system calculates all the... Among (i=1 to M), find the smallest value and denote it as . And record its corresponding standard fault feature matrix. .

[0074] This means that, mathematically speaking, the current mode is most similar to the k-th failure mode.

[0075] Then, in sub-step S43, the system does not directly adopt this result, but further determines the minimum distance. Is it less than a preset matching threshold? This threshold acts as a "confidence threshold." If If the distance is less than the threshold, it means that the current pattern is sufficiently similar to the standard pattern C_k, and the match is successful. Conversely, if the distance is still too large, even if it is the "most similar", it may mean that the current pattern does not belong to any known typical fault, or the system is in normal fluctuation. In this case, it may not trigger an alert or trigger an "unknown anomaly" prompt.

[0076] By concretizing the "matching degree comparison" process into an operable logical chain of "distance calculation - finding the minimum - threshold judgment," the entire pattern recognition process becomes clear, rigorous, and controllable. Finding the minimum distance ensures that the system always attempts to provide the most probable diagnosis, while setting a matching threshold increases the system's robustness, preventing forced matching to a fault mode when the system only experiences slight, meaningless random fluctuations, thus effectively reducing false positives. This design improves early warning sensitivity while ensuring early warning accuracy, providing a crucial logical guarantee for achieving intelligent and reliable fault diagnosis.

[0077] Specifically, step S5, "generating and outputting early warning suggestion information based on the matching degree comparison results," includes: S51. Determine whether the minimum matrix distance is less than a preset matching threshold; S52. When the determination is yes, the device anomaly mode associated with the standard fault feature matrix corresponding to the minimum matrix distance is determined as the currently identified target fault mode. S53. Generate and output early warning suggestion information, wherein the early warning suggestion information includes a textual description of the target fault mode and operation suggestions corresponding to the target fault mode.

[0078] In this embodiment, the "equipment abnormality modes" associated with the pre-stored standard fault feature matrix library are preferably several specific fault types that are most common in clinical practice and urgently need to be distinguished, including but not limited to at least one of "filter coagulation mode", "venous reservoir fluid level abnormality mode" and "arterial end drainage obstruction mode". These modes have distinctly different pathophysiological mechanisms and pressure parameter change characteristics.

[0079] Accordingly, when generating the early warning suggestion information in step S5, the indicated "equipment abnormal mode" is: the standard fault feature matrix with the smallest matrix distance to the current real-time risk covariance matrix C(t) in the matching degree comparison. The specific fault mode associated with the fault, and the distance difference is below a threshold. For example, if the calculation shows that C(t) is at its minimum distance from the standard fault feature matrix representing "filter clotting mode" and is below the threshold, the warning suggestion will indicate "filter clotting mode". The system output will no longer be a vague "pressure anomaly", but will point to a specific, actionable fault point.

[0080] The following are some examples: (1) Example 1: Early stage of filter coagulation Traditional alarms: "High transmembrane pressure alarm" or "High venous pressure alarm".

[0081] The blood filtration machine provided by this invention displays warnings and provides operation suggestions on the screen: ① Warning: Early coagulation pattern detected in the filter (92% match). Characteristics: Transmembrane pressure and venous pressure show a continuous, coordinated, and slow upward trend (strong positive correlation of covariance), and the cumulative risk score of transmembrane pressure has reached 75.

[0082] ②Recommendations: 1. Immediately check the color of the filter fibers; 2. Consider increasing the anticoagulant pump rate from the [current value] to the [recommended value] ml / h; 3. Prepare for a prophylactic saline flush. The fault is located inside the filter (blood filter). The instructions directly address anticoagulant adjustment and filter management.

[0083] (2) Example 2: The fluid level in the venous infusion chamber is too high or there is blood clotting. Traditional alarm: "High venous pressure alarm".

[0084] The blood filtration machine provided by this invention displays warnings and provides operation suggestions on the screen: ① Warning: An abnormal pattern in the venous ampulla was detected (match rate 85%). Characteristics: Venous pressure shows a step increase followed by sustained high-level fluctuations, with a significantly reduced correlation with changes in arterial pressure, and a sharp increase in the variance of venous pressure fluctuations.

[0085] ②Recommendations: 1. Immediately check if the fluid level in the venous infusion chamber is too low or if there are blood clots; 2. If the fluid level is too low, adjust the fluid level to the safe level; 3. If blood clots are found, treat the venous infusion chamber according to the procedure.

[0086] The identified fault location is the vein chamber (air trapping chamber). The procedure directly addresses the inspection and repair of the vein chamber.

[0087] (3) Example 3: Poor drainage at the arterial end (e.g., catheter adhering to the wall, partial blockage) Traditional alarms: "Low arterial pressure alarm" or "Low blood flow rate alarm".

[0088] The blood filtration machine provided by this invention displays warnings and provides operation suggestions on the screen: ① Warning: A pattern of poor arterial drainage was detected (match rate 88%). Characteristics: The absolute value of arterial pressure is low and exhibits sharp, sawtooth-like fluctuations (large variance), with decreased synergy with blood flow velocity. Simultaneously, changes in arterial pressure are decoupled from changes in venous pressure.

[0089] ②Recommendations: 1. Check the patient's catheter position and fixation; 2. Try adjusting the patient's position; 3. Gently flush or adjust the catheter depth, do not force aspiration. The fault location is indicated at the tip of the arterial catheter within the patient's blood vessel. The procedure directly addresses the catheter position and patient positioning.

[0090] (4) Example 4: Pipeline kink or pressure Traditional alarms: "High venous pressure alarm" or "Pressure limit alarm".

[0091] The blood filtration machine provided by this invention displays warnings and provides operation suggestions on the screen: ① Warning: Cardiopulmonary bypass tubing obstruction pattern detected (90% match). Characteristics: Sudden and significant increase in venous pressure (step change), while transmembrane pressure changes are not significant, and the ratio of arteriovenous pressure gradient to blood flow velocity increases instantaneously.

[0092] ②Recommendations: 1. Immediately inspect the entire extracorporeal circulation tubing, especially at bends and points of pressure risk; 2. Remove any kinks or compression; 3. Confirm tubing patency. The identified fault is a physical obstruction in the extracorporeal circulation tubing (especially the venous circuit). The procedure should directly involve checking the tubing.

[0093] In summary, the early warning and suggestion information output of this invention, through pattern recognition, achieves a three-tiered leap from "what's wrong" (high pressure) to "where the problem is likely" (filter / intravenous chamber / catheter / line), and then to "what should be done first" (targeted examination and treatment). This significantly reduces the mental burden on medical staff, shortens the path from alarm to effective intervention, and allows precious rescue time to be spent directly solving the problem rather than troubleshooting. This specific and actionable fault point identification is precisely the core value of this solution, surpassing existing technologies and providing "intelligent decision support."

[0094] On the one hand, it concretizes the abstract "abnormal patterns" into typical clinical failure scenarios familiar to medical personnel, making the application purpose and the practical problems it solves extremely clear. On the other hand, it clarifies the output logic of early warning and suggestion information, ensuring the directness of the information and its action guidance.

[0095] Medical staff no longer receive raw data or simple alarms that require time to interpret, but rather preliminary diagnostic conclusions after intelligent system analysis, such as "suspected early clotting of the filter" or "pay attention to checking the fluid level in the venous reservoir." This enables them to take the most targeted measures immediately, greatly improving the clinical utility and value of alarm information, fundamentally changing the efficiency and effectiveness of human-computer interaction, and achieving a leap from "noise alarm" to "decision support."

[0096] In summary, the solution provided in this embodiment has the following advantages: ① By calculating the covariance matrix of the risk state vector and analyzing its matching degree with the standard fault feature matrix, a leap from "single parameter threshold judgment" to "multi-parameter dynamic collaborative pattern recognition" has been achieved, solving the fundamental problems of delayed early warning and high false alarm rate in traditional alarm mechanisms.

[0097] ② By matching the real-time calculated risk covariance matrix with a variety of pre-stored fault feature matrices, specific early fault patterns (such as filter coagulation and tubing abnormalities) can be identified before the parameters exceed the static threshold, achieving true early warning and buying valuable time for clinical intervention.

[0098] ③ Based on the matrix matching results, it directly outputs early warning information containing specific fault mode descriptions (such as "early coagulation of the filter"), providing intelligent root cause analysis and preliminary diagnosis functions that are lacking in traditional alarms, significantly reducing the mental burden and investigation time of medical staff.

[0099] ④ By associating and outputting specific fault modes with corresponding operational suggestions (such as "check the filter" and "adjust the anticoagulant rate"), a leap from "alarm" to "decision support" is achieved, directly improving the pertinence and efficiency of clinical treatment.

[0100] ⑤ The constructed risk covariance matrix includes the fluctuation intensity (variance) of each parameter and the cooperative relationship (covariance) between them, providing a comprehensive and quantitative feature description for fault mode matching, making the judgment of the entire early warning system more robust and accurate.

[0101] Example 2 This embodiment aims to break down the entire process of inputting, calculating, and outputting the risk covariance matrix in the blood filtration equipment early warning method based on multi-parameter coupling provided in Embodiment 1, in conjunction with actual monitoring scenarios.

[0102] I. Core Application Prerequisite: Clearly Define the Source of Input Data 1. Basic data: Time-series data of risk status indicators collected in real time by the sensors of the blood filtration machine, including at least transmembrane pressure (TMP) and venous pressure (PV), and may also include arterial pressure (PA) and blood flow velocity (QB) (corresponding to step S1); 2. Intermediate data: Risk state vectors constructed based on basic data. ,in , These are risk status indicators, for example: = Transmembrane pressure change rate, =Instantaneous venous pressure value = Risk integral function F1(t); 3. Input sample set: Set time window (For example, W=5 minutes, which can be adjusted according to the device's response accuracy), continuously collect k risk state vector samples within this window to form a sample sequence: (k is the number of samples within the window; for example, if the sample is taken once per second, then k = 300).

[0103] II. Core Calculation Process: Scenario-based Derivation of the Risk Covariance Matrix The calculation of the "risk covariance matrix C(t)" in step S3 needs to be combined with the actual indicator characteristics of the blood filtration machine monitoring scenario. Taking "2 core risk indicators" as an example (for ease of understanding, it can be extended to n), the complete derivation process is as follows: 1. Define scenario-based variables: Assume the risk state vector is... ,in: = Rate of change of transmembrane pressure (TMP) (Risk indicator 1, directly reflecting the filter clogging trend); 2. = Rate of change in venous pressure (PV) (Risk indicator 2, directly reflecting the circulation status of the circuit).

[0104] 3. Matrix Dimension Definition: Since the vector dimension is 2, the risk covariance matrix... It is a 2×2 symmetric matrix, in the following form:

[0105] 4. Contextual meaning and calculation of matrix elements: Diagonal elements (self-volatility): represent the fluctuation range of a single risk indicator within a time window, reflecting the stability of the indicator itself. The variance of the rate of change of transmembrane pressure; a large value indicates that the transmembrane pressure fluctuates violently (which may be a precursor to filter clotting). 5. Calculation logic: ,in Let m be the rate of change of transmembrane pressure for the m-th sample. This represents the average rate of change of transmembrane pressure within the window.

[0106] 6. Off-diagonal elements (degree of synergistic change): represent the linkage between two risk indicators and are the core quantitative manifestation of "multi-parameter coupling". The covariance between the rate of change of transmembrane pressure and the rate of change of venous pressure indicates that the two are correlated and rise and fall together (e.g., when the filter clots, both rise synchronously). A value close to 0 indicates that the two are not correlated (e.g., when the fluid level in the venous chamber is abnormal, only the venous pressure fluctuates). 7. Calculation Logic: ,in Let m be the rate of change of venous pressure for the m-th sample. This represents the average rate of change in venous pressure within the window.

[0107] III. Core Output Value: Matching Logic with Fault Modes (Connecting to Step S4) 1. Pre-stored standard fault feature matrix: Extract standard fault feature matrices corresponding to different equipment anomaly modes through a large amount of experimental data. ,For example: Filter Coagulation Mode Template : (Transmembrane pressure fluctuations) and (Venous pressure fluctuations) were all significantly greater than the normal threshold, and A relatively large positive value (both increase simultaneously); Abnormal fluid level pattern template for venous urn : (Venous pressure fluctuations) are much greater than the normal threshold. Approaching the normal threshold, and (The two are unrelated).

[0108] 2. Real-time matching Calculate the real-time risk covariance matrix With each standard fault characteristic matrix distance :

[0109] in, Real-time risk covariance matrix The element in row p and column q; Standard Fault Feature Matrix The element in row p and column q; n: Matrix dimension (equal to the number of risk status indicators).

[0110] 3. Warning Triggered: If the real-time matrix With filter coagulation mode template distance Minimum and Less than the matching threshold Output "Filter clotting warning"; If real-time matrix Abnormal fluid level pattern template for venous urinal distance Minimum and Less than the matching threshold Output "Abnormal fluid level warning in the venous reservoir".

[0111] Example 3 This embodiment provides a blood filtration machine for executing any of the blood filtration equipment early warning methods based on multi-parameter coupling provided in Embodiment 1, and has the same function and beneficial effects.

[0112] See Figure 2 The blood filtration machine provided in this embodiment includes: The data monitoring module 1 is used to acquire time-series data of at least two risk status indicators of the blood filtration machine in real time, wherein the at least two risk status indicators include transmembrane pressure and venous pressure; Vector construction module 2 is used to construct a risk state vector based on the time-series data of the at least two risk state indicators; Matrix calculation module 3 is used to calculate a risk covariance matrix based on the change of the risk state vector within a set time window, wherein the risk covariance matrix is ​​used to characterize the synergistic relationship between the changing trends of the at least two risk state indicators. Matching module 4 is used to compare the calculated risk covariance matrix with at least one pre-stored standard fault feature matrix, wherein different standard fault feature matrices are associated with different equipment anomaly modes. The early warning suggestion module 5 is used to generate and output early warning suggestion information based on the matching degree comparison results; wherein, the early warning suggestion information indicates at least one abnormal device mode and corresponding operation suggestions.

[0113] Based on Example 1, features not explained in this example will be explained using the methods described in Example 1, and will not be repeated here.

[0114] Finally, it should be noted that although the above embodiments have been described in the text and drawings of this application, this should not limit the scope of patent protection of this application. Any technical solutions that are based on the essential concept of this application and utilize the content described in the text and drawings of this application, resulting in equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this application.

Claims

1. A method for early warning of blood filtration equipment based on multi-parameter coupling, characterized in that, include: S1. Real-time acquisition of time-series data of at least two risk status indicators of the blood filtration machine operation, wherein the at least two risk status indicators include transmembrane pressure and venous pressure; S2. Construct a risk state vector based on the time-series data of the at least two risk state indicators; S3. Based on the changes of the risk state vector within a set time window, calculate a risk covariance matrix, wherein the risk covariance matrix is ​​used to characterize the synergistic relationship between the changing trends of the at least two risk state indicators. S4. Compare the calculated risk covariance matrix with at least one pre-stored standard fault feature matrix to determine the matching degree, wherein different standard fault feature matrices are associated with different equipment anomaly modes. S5. Based on the matching degree comparison result, generate and output early warning suggestion information; wherein, the early warning suggestion information indicates at least one abnormal device mode and corresponding operation suggestions.

2. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 1, characterized in that, In step S2, the constructed risk state vector R(t) is expressed as: in, n: a positive integer not less than 2; [x1(t), x2(t), ..., x n [(t)]: The values ​​of n risk status indicators determined based on the time series data of the at least two risk status indicators at time t; [x1(t), x2(t), ..., x n (t)] T : represents [x1(t), x2(t), ..., x n The transpose of [(t)].

3. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 2, characterized in that, The n risk status indicators include at least two of the following: the instantaneous value of the transmembrane pressure, the rate of change of the transmembrane pressure, the instantaneous value of the venous pressure, the rate of change of the venous pressure, a risk integral function value constructed based on the transmembrane pressure, a risk integral function value constructed based on the venous pressure and arterial pressure, and the rate of change of the pressure difference between the venous pressure and the arterial pressure.

4. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 3, characterized in that, The risk integral function F1(t) constructed based on the transmembrane pressure is defined as follows: in, T is the start time of integration, and t is the end time of integration; Transmembrane pressure as a function of time A changing function; for Rate of change per unit time; This is a mapping function used to convert the rate of change of transmembrane pressure. This is mapped to a risk contribution rate for accumulation.

5. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 3, characterized in that, The risk integral function F2(t) constructed based on the venous pressure and arterial pressure is defined as follows: in, T is the start time of integration, and t is the end time of integration; This refers to the pressure difference between the venous and arterial ends of the extracorporeal circulation loop. Blood flow velocity; This is a mapping function used to represent the rate of change of flow resistance. This is mapped to a risk contribution rate for accumulation.

6. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 1, characterized in that, Step S3 includes: Define a continuous historical time period, i.e., a time window [tW, t], where t is the current time and W is the preset window length; Obtain a series of values ​​for the risk state vector R(τ) within the time window [tW, t], a total of N observation samples arranged in chronological order, denoted as R(1), R(2), …, R(N); where each observation sample R(k) (k=1,2,...,N) is an n-dimensional column vector. , This represents the value of the i-th risk status index at the k-th sampling time; Construct an n × n risk covariance matrix C(t) based on these N samples; The steps for constructing the risk covariance matrix C(t) include: First, calculate the sample mean of each risk status indicator within the time window. : ; Then, each element in the risk covariance matrix C(t) is calculated. ;in, The covariance between the i-th and j-th risk status indicators within the time window is represented by the following formula: Where; when i = j, the calculated This is the sample variance of the i-th risk status indicator within that time window: ; All calculated elements This forms an n × n real symmetric matrix, namely the risk covariance matrix C(t).

7. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 1, characterized in that, Step S4 specifically includes: S41. Calculate the risk covariance matrix respectively. Matrix distance between each of the pre-stored standard fault feature matrices ,in, M is the total number of standard fault feature matrices; where different standard fault feature matrices are associated with different equipment anomaly modes. S42. Determine the minimum matrix distance. and its corresponding standard fault feature matrix ; S43. Determine the minimum matrix distance. Is it less than the preset matching threshold? 8. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 7, characterized in that, In step S41, in, Real-time risk covariance matrix The element in row p and column q; Standard Fault Feature Matrix The element in row p and column q; n: Matrix dimension, equal to the number of risk status indicators.

9. The early warning method for blood filtration equipment based on multi-parameter coupling according to claim 7 or 8, characterized in that, Step S5 includes: S51. Determine whether the minimum matrix distance is less than a preset matching threshold; S52. When the determination is yes, the device anomaly mode associated with the standard fault feature matrix corresponding to the minimum matrix distance is determined as the currently identified target fault mode. S53. Generate and output early warning suggestion information, wherein the early warning suggestion information includes a textual description of the target fault mode and operation suggestions corresponding to the target fault mode.

10. A blood filtration machine, used to execute the blood filtration equipment early warning method based on multi-parameter coupling as described in claims 1-9, characterized in that, include: The data monitoring module is used to acquire time-series data of at least two risk status indicators of the blood filtration machine in real time, wherein the at least two risk status indicators include transmembrane pressure and venous pressure; A vector construction module is used to construct a risk state vector based on the time-series data of the at least two risk state indicators. The matrix calculation module is used to calculate a risk covariance matrix based on the changes of the risk state vector within a set time window, wherein the risk covariance matrix is ​​used to characterize the synergistic relationship between the changing trends of the at least two risk state indicators. The matching module is used to compare the calculated risk covariance matrix with at least one pre-stored standard fault feature matrix, wherein different standard fault feature matrices are associated with different equipment anomaly modes. The early warning suggestion module is used to generate and output early warning suggestion information based on the matching degree comparison results; wherein, the early warning suggestion information indicates at least one abnormal device mode and corresponding operation suggestions.

Images