Multi-dimensional dynamic trend index processing method and system

Abstract

The application discloses a multi-dimensional dynamic situation index processing method and system, and relates to the technical field of data processing. The method comprises the following steps: acquiring multi-dimensional dynamic index data and object attributes of a monitoring object, calling a matched physical constraint model, identifying driving indexes, response indexes and their evolution relationship with a physical driving relationship; predicting the current response value according to historical data, calculating the degree of physical consistency deviation, and judging stability in combination with the distance and change speed of the indexes from the safety boundary. When stable, a low resource mode is adopted, otherwise a complete mode is adopted; when the mode is switched, the dynamic situation result is generated by fusing the historical data before the anomaly and the current data. According to the deviation degree of the actual response relative to the theoretical response, the application can identify the situation that the indexes have not crossed the boundary but the physical relationship between the indexes has been abnormal; the physical evolution relationship is used to capture the associated abnormality, so that the effect of discovering potential faults, performance degradation or state instability trend in advance is achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for processing multi-dimensional dynamic situation indicators. Background Technology

[0002] With the development of technologies such as industrial equipment operation monitoring, energy system status analysis, environmental situation awareness, and complex infrastructure safety management, it is usually necessary to continuously collect dynamic indicator data of the monitored objects in multiple dimensions, and to generate dynamic situation results that reflect the current operating status and changing trends of the monitored objects by analyzing the dynamic indicator data.

[0003] Existing multidimensional dynamic situational awareness processing methods typically rely on preset thresholds and changes as the primary criteria for judgment. For example, they may determine whether each dynamic indicator exceeds its corresponding threshold, or determine whether the monitored object is in a stable state based on the rate of change of indicator values ​​within adjacent sampling periods. When indicators are within limits and the changes are small, the monitored object is usually considered to be operating stably; when indicators approach or exceed safety boundaries, the monitored object is considered to have an abnormal risk.

[0004] However, for monitoring objects with clear physical operating mechanisms, different dynamic indicators are not independent of each other. Instead, they may represent the driving factors acting on the monitoring object and the state response caused by the driving factors, respectively. A physical evolutionary relationship exists between the two, determined by the object's structure, type, and operating conditions. In the early stages of anomaly formation, the response indicators may not yet exceed the preset safety boundary, and their rate of change may be within the normal range. However, the change in the response indicators relative to the driving indicators has already deviated from the normal physical evolutionary pattern. At this point, if judgment is still made solely based on indicator values, rate of change, or distance from the safety boundary, it is easy to mistake a state that has actually undergone abnormal evolution for a stable state.

[0005] The aforementioned deviations in stable state identification directly impact subsequent dynamic situation processing results. When the system reduces data processing intensity based on inaccurate stability assessments, the early changes in physical anomalies may not be fully analyzed, causing the generated dynamic situation results to fail to accurately reflect the state evolution of the monitored object. Conversely, consistently employing high-intensity processing to avoid missing anomalies results in a large amount of data in a truly stable state continuously consuming computational, storage, and transmission resources. Therefore, existing methods for stability assessment based on surface numerical characteristics of indicators cannot simultaneously meet the requirements of both the reliability of dynamic situation results and the efficiency of processing resource utilization. Summary of the Invention

[0006] To address the problem that existing technologies struggle to balance reliability and resource utilization efficiency, this invention provides a method and system for processing multi-dimensional dynamic situation indicators.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] Firstly, this application discloses a method for processing multi-dimensional dynamic situation indicators, including the following steps:

[0009] Obtain multidimensional dynamic indicator data and object attributes corresponding to the monitored object;

[0010] Based on object attributes, the physical constraint model matching the monitored object is retrieved from the model database. Based on the physical constraint model, the driving indicator data and response indicator data with physical driving relationship and the physical evolution relationship between the two are distinguished from the multidimensional dynamic indicator data.

[0011] The predicted value of the current response index is calculated based on the physical evolution relationship, driving index data, and response index data, and the deviation from the currently acquired response index data is calculated to obtain the degree of physical consistency deviation.

[0012] Calculate the distance between the multidimensional dynamic index data and the preset safety boundary, as well as the rate of change of the dynamic index, and combine the degree of deviation from physical consistency to determine whether the stability condition is met;

[0013] If the conditions are met, a low-resource processing mode is adopted for the multidimensional dynamic indicator data and dynamic situation results are generated.

[0014] If the conditions are not met, the full processing mode is adopted and a dynamic situational result is generated. If the current process is switching from the low-resource processing mode to the full processing mode, the historical dynamic indicator data within the preset time range before the anomaly occurred is called and fused with the current multi-dimensional dynamic indicator data to generate a dynamic situational result.

[0015] Secondly, this application discloses a system for processing multi-dimensional dynamic situation indicators, including:

[0016] The data acquisition module is used to acquire multidimensional dynamic indicator data and object attributes corresponding to the monitored object;

[0017] The model invocation module is used to retrieve physical constraint models that match the monitored objects from the model database based on object attributes. Based on the physical constraint models, it distinguishes driving indicator data and response indicator data with physical driving relationships and the physical evolution relationship between the two from multidimensional dynamic indicator data.

[0018] The deviation calculation module is used to calculate the current predicted value of the response index based on the physical evolution relationship, driving index data and response index data, and to calculate the degree of deviation of the physical consistency with the currently acquired response index data.

[0019] The output module is used to calculate the distance between the multi-dimensional dynamic indicator data and the preset safety boundary, as well as the rate of change of the dynamic indicators. It combines the degree of deviation from physical consistency to determine whether the stability condition is met, and generates dynamic situation results in the following manner:

[0020] If the conditions are met, a low-resource processing mode is adopted for the multidimensional dynamic indicator data and dynamic situation results are generated.

[0021] If the conditions are not met, the full processing mode is adopted and a dynamic situational result is generated. If the current process is switching from the low-resource processing mode to the full processing mode, the historical dynamic indicator data within the preset time range before the anomaly occurred is called and fused with the current multi-dimensional dynamic indicator data to generate a dynamic situational result.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] This application can identify situations where the physical relationship between indicators is abnormal even though the indicators have not yet exceeded the limits by measuring the deviation of the actual response from the theoretical response; by using the physical evolution relationship to capture such correlation anomalies, it has the effect of detecting potential faults, performance degradation or state instability trends in advance.

[0024] Based on this, this application dynamically selects between a low-resource processing mode and a full processing mode according to the stability assessment results, thereby reducing redundant computations in a stable state while still enabling complete analysis in anomaly states. Simultaneously, when switching from the low-resource processing mode to the full processing mode, historical dynamic indicator data prior to the anomaly is retrospectively reviewed and integrated to complete the anomaly formation process, avoiding gaps in situational information caused by simplified resource processing. This balances the reliability, completeness, and resource utilization efficiency of dynamic situational processing. Attached Figure Description

[0025] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0026] Figure 1 This is a flowchart of the multi-dimensional dynamic situation indicator processing method introduced in this invention;

[0027] Figure 2 For based on Figure 1 Application scenarios;

[0028] Figure 3 A flowchart of a system for processing multi-dimensional dynamic situation indicators;

[0029] Figure 4 For based on Figure 3 The system block diagram with an anomaly location module. Detailed Implementation

[0030] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0031] Example 1

[0032] like Figure 1 and Figure 2 This paper describes a method for processing multi-dimensional dynamic situation indicators, applied to scenarios where an industrial internet platform performs multi-dimensional dynamic situation monitoring of monitored objects. It should be understood that this method can be executed by the industrial internet platform, which may include multiple data acquisition devices, edge processing nodes, a communication network, and a cloud data processing platform located on the monitored object side. Specifically, the data acquisition devices collect multi-dimensional dynamic indicator data during the operation of the monitored object; the edge processing nodes receive the dynamic indicator data generated by the data acquisition devices and perform data caching, summary data calculation, and data uploading according to the current processing mode; the cloud data processing platform may include a model database, a model calling module, a deviation calculation module, a judgment output module, and an anomaly location module. The model database pre-stores physical constraint models matching different types of monitored objects. The specific hardware configuration of the above devices and their communication methods can be implemented using conventional methods, and this embodiment of the invention is not limited thereto.

[0033] The following example illustrates this method using an industrial internet platform for multi-dimensional dynamic situation monitoring of motor equipment. It should be understood that the monitoring object can also be a pump unit, compressor unit, power transmission line, pipeline, or remote sensing target, etc., and this embodiment of the invention is not limited to these.

[0034] like Figure 1 As shown, it includes the following steps:

[0035] S101. Obtain multi-dimensional dynamic indicator data and object attributes corresponding to the monitored object.

[0036] Specifically, the edge processing node can acquire multi-dimensional dynamic index data of the motor device according to a preset local sampling period using current sensors, voltage sensors, temperature sensors, vibration sensors, speed sensors, torque sensors, and ambient temperature sensors. In this embodiment of the invention, the multi-dimensional dynamic index data includes motor operating current, motor operating voltage, motor input power, motor speed, motor winding temperature, motor vibration amplitude, and ambient temperature. It should be understood that some of the above-mentioned multi-dimensional dynamic index data can be directly acquired by the data acquisition device, or can be calculated from the directly acquired indexes. For example, the motor input power can be calculated from the motor operating current and the motor operating voltage. This embodiment of the invention is not limited to this.

[0037] Let the time interval between two adjacent processing cycles be . Then the first The time corresponding to each processing cycle is In the formula, This indicates the processing cycle length, which should be matched to the response speed of the physical constraint model being invoked, preferably between 10 and 120 seconds. Indicates the processing cycle number; Indicates the start time. The edge processing node at time [time missing]. The obtained multidimensional dynamic indicator data is denoted as .

[0038] Optionally, the local sampling period is preferably between 100 milliseconds and 2 seconds. The local sampling period should meet the sampling requirements for the fastest physical process to be observed in the monitored object. For motor equipment, the time constant of the electrical process is on the order of milliseconds, and the time constant of the mechanical process is on the order of hundreds of milliseconds.

[0039] Optionally, the object attributes may include motor type, motor model, rated power, rated current, rated speed, winding resistance, equivalent heat capacity, equivalent thermal resistance, moment of inertia, damping coefficient, cooling method, and installation environment. These object attributes can be obtained through device registration information, device configuration files, or a device management platform; however, this embodiment of the invention is not limited to these methods.

[0040] It should be noted that the local sampling frequency of the data acquisition device does not need to be reduced with the change of processing mode. Even if the system enters low-resource processing mode, the edge processing node can still collect raw dynamic indicator data at a frequency that meets the requirements of device safety monitoring, and only adjust the data upload frequency, cloud storage frequency and data processing form, so as to ensure that when it is necessary to restore the anomaly formation process, the edge side always retains raw dynamic indicator data with sufficient granularity. For details of this low-resource processing mode, please refer to S105 below.

[0041] S102. Based on the object attributes, call the physical constraint model matching the monitored object from the model database, and distinguish the driving indicator data and response indicator data with physical driving relationship and the physical evolution relationship between them from the multidimensional dynamic indicator data according to the physical constraint model.

[0042] The physical constraint models include thermal state constraint models based on energy transfer relationships, motion state constraint models based on dynamic relationships, electrical state constraint models based on electrical characteristics, flow state constraint models based on fluid transport relationships, and remote sensing target state constraint models based on spatial variation relationships.

[0043] Physical evolution relationships include the time delay relationship in which driving index data affects response index data, as well as the mapping relationship of state changes;

[0044] The start time of the preset time range for historical dynamic indicator data is determined based on the time delay relationship.

[0045] It should be understood that this physical constraint model is a model pre-established and stored in a model database based on the physical laws followed by the monitored object. It records the physical role of each dynamic indicator in the corresponding physical process, the transmission relationship between dynamic indicators, the state change mapping relationship, and the time delay relationship. The above five types of physical constraint models correspond to different types of monitored objects: the thermal state constraint model describes the heat generation and dissipation process of the monitored object; the motion state constraint model describes the interaction between torque and motion state; the electrical state constraint model describes the constraint process between voltage, current, and power; the flow state constraint model describes the transmission process between pressure, flow rate, and resistance; and the remote sensing target state constraint model describes the evolution process between the target's position, velocity, and spatial state.

[0046] Specifically, the cloud-based data processing platform can retrieve a physical constraint model matching the current motor device from the model database based on the motor type, motor model, rated power, winding resistance, equivalent heat capacity, equivalent thermal resistance, and cooling method obtained in S101. For the motor device in this embodiment of the invention, thermal state constraint model, electrical state constraint model, and motion state constraint model can be retrieved.

[0047] For the thermal state constraint model, the thermal balance relationship during motor operation can be expressed as:

[0048] ;

[0049] In the formula, Indicates the equivalent heat capacity of the motor; Indicates the equivalent thermal resistance of the motor; Indicates the time when the motor is Power loss; This represents the rate of change of winding temperature over time. Indicates the temperature of the motor windings; This represents the amplitude of motor vibration. The left side of the formula represents the heat accumulated in the winding per unit time, the first term on the right side represents the heat generated by losses, and the second term on the right side represents the heat transferred to the environment through the heat dissipation path.

[0050] Optionally, this power loss can be taken as the motor copper loss, that is, the square of the motor operating current and the equivalent resistance of the motor windings. The product of the losses and the losses. In another embodiment, the total power loss of the motor can be determined by further combining iron loss, mechanical loss, and additional loss, but the embodiments of the present invention are not limited thereto.

[0051] For the motion state constraint model, the rotation state of the motor satisfies:

[0052] ;

[0053] In the formula, This represents the moment of inertia of the motor. Indicates the time when the motor is angular velocity; Indicates electromagnetic torque; Indicates load torque; This represents the equivalent damping coefficient. The relationship between angular velocity and rotational speed is such that angular velocity equals rotational speed. The product of and 60; the electromagnetic torque can be taken as the motor torque constant. The product of the load torque and the operating current. Optionally, the load torque can be directly acquired by a torque sensor, or it can be obtained by inversely calculating the electromagnetic torque and angular velocity according to the above-mentioned motion state constraint model when the motor is in steady state operation. The embodiments of the present invention are not limited thereto.

[0054] Furthermore, the vibration response relationship is also recorded in the motion state constraint model. When the equivalent rotor imbalance is constant, the vibration amplitude of the motor is proportional to the square of the angular velocity, i.e.:

[0055] ;

[0056] In the formula, This represents the predicted amplitude of the vibration during the current period. The value of represents the vibration response coefficient, which can be obtained by calibrating the vibration amplitude and angular velocity of the monitored object under normal operating conditions; the angular velocity in the above formula is the predicted value of the current cycle angular velocity obtained by recursion from the data of the previous cycle. This represents the square of the predicted angular velocity value for the current period.

[0057] Accordingly, in this vibration response relationship, rotational speed is determined as the driving index data, and vibration amplitude is determined as the response index data. Therefore, the physical evolution relationship corresponding to the motor vibration state can be expressed as:

[0058] ;

[0059] Indicates the motor speed; express The amplitude of motor vibration at any given time. In this embodiment of the invention, driving index data refers to index data that, in the corresponding physical process, acts as energy input, motion input, environmental input, or state excitation, and has a physical impact on other dynamic indicators; the response index data refers to index data that is affected by the driving index and is used to reflect the state result of the monitored object. The cloud data processing platform can distinguish between driving index data and response index data from multidimensional dynamic index data based on the pre-stored physical roles of the indicators in the physical constraint model.

[0060] Specifically, in the thermal state constraint model, the operating current, power loss, and ambient temperature are determined as driving index data, and the winding temperature is determined as the response index data. The corresponding physical evolution relationship can be expressed as:

[0061] ;

[0062] Indicates the motor's operating current; Indicates ambient temperature; Indicates the time when the motor is Power loss; express The temperature of the motor windings at any given time.

[0063] In the motion state constraint model, the operating current and load torque are determined as driving index data, and the rotational speed is determined as response index data. The corresponding physical evolution relationship can be expressed as:

[0064] ;

[0065] Indicates the motor's operating current; Indicates load torque; express The motor speed at any given time.

[0066] For the electrical state constraint model, the input power of the motor satisfies the following relationship with the operating voltage and operating current:

[0067] ;

[0068] In the formula, Indicates that the motor is Input power at any given time; Indicates the motor's operating voltage; The motor power factor can be obtained from object attributes or from normal operation data calibration. Accordingly, in this power state constraint model, operating voltage and operating current are determined as driving index data, and input power is determined as response index data. The corresponding physical evolution relationship can be expressed as:

[0069] ;

[0070] It should be noted that the time constant of the electrical process is on the order of milliseconds, which is much smaller than the processing cycle length. Therefore, the time delay parameter corresponding to this physical evolution relationship can be taken to be within 1 second. When calculating the predicted value of this response index, the driving index data of the current cycle is used.

[0071] For the flow state constraint model, the inlet and outlet pressures and flow rates during pipeline transportation satisfy the friction resistance relationship:

[0072] ;

[0073] In the formula, , These respectively indicate the pipeline at Constant pressure on imports and exports; Indicates flow rate; This represents the friction factor, the value of which can be determined by pipeline design parameters or obtained from normal operating data calibration. Accordingly, in this flow state constraint model, pump speed and inlet pressure are defined as driving index data, and flow rate and outlet pressure are defined as response index data. The corresponding physical evolution relationship can be expressed as:

[0074] ;

[0075] In the formula, Indicates the pump speed; express Flow rate at any given moment; express The outlet pressure at a given moment. The time delay parameter corresponding to this physical evolution relationship depends on the ratio of pipe length to fluid velocity, and is typically on the order of seconds to tens of seconds.

[0076] For the remote sensing target state constraint model, the target position and target velocity satisfy a kinematic relationship:

[0077] ;

[0078] In the formula, This represents the predicted target location for the current period. Indicates the target position of the previous period; This indicates the target speed for the previous cycle; This indicates the processing cycle length. Accordingly, in this remote sensing target state constraint model, target velocity and target acceleration are determined as driving index data, and target position is determined as response index data. The corresponding physical evolution relationship can be expressed as:

[0079] ;

[0080] In the formula, Indicates the target acceleration; express The target location at any given time. The time delay parameter corresponding to this physical evolution relationship is determined by the observation period. The calling methods of the above-mentioned power state constraint model, flow state constraint model, and remote sensing target state constraint model, as well as the methods of distinguishing between driving index data and response index data, are the same as those of the aforementioned thermal state constraint model and motion state constraint model, and the embodiments of the present invention are not limited thereto.

[0081] It should also be noted that the same dynamic index can play different physical roles in different physical evolution relationships. For state indices located in the middle of the physical evolution path, they are response indices relative to upstream indices, and driving indices relative to downstream indices. For example, power loss is a response index relative to operating current, but a driving index relative to winding temperature; rotational speed is a response index relative to electromagnetic torque, but a driving index relative to vibration amplitude. For response indices located at the end of the physical evolution path, since the state change mapping relationship is a recursive form across cycles, the actual collected value of the previous cycle constitutes one of the inputs for calculating the predicted value of the response index in the current cycle. For example, the actual collected value of winding temperature in the previous cycle is one of the inputs for calculating the predicted value of winding temperature in the current cycle.

[0082] It's important to note that the distinction between driving and response indicator data stems from the physical direction of action determined by the physical constraint model, not from the ordinary statistical correlation between multiple dynamic indicators. Therefore, even if two dynamic indicators exhibit strong synchronous fluctuations in value, they will not be identified as indicators with a physical driving relationship if no physical interaction exists within the physical constraint model. Conversely, even if two dynamic indicators have low statistical correlation at the same moment due to time delay, they will still be identified as indicators with a physical driving relationship if a physical interaction exists within the physical constraint model. This avoids establishing data relationships lacking physical basis.

[0083] For example, in a specific application scenario, the cooling fan of a motor experiences reduced ventilation due to dust accumulation from long-term operation. This reduces the motor's equivalent heat dissipation capacity, but the operating current, operating voltage, speed, and load remain relatively stable, and the ambient temperature remains stable. Only the winding temperature rises slowly at a low rate. In this scenario, if a statistical correlation-based approach is used, the correlation coefficient between current and temperature is very low because the current data does not fluctuate significantly during this period, making it impossible to establish a relationship between the two. Furthermore, since the winding temperature is still within a safe range, alarms based on single-threshold parameters will not be triggered, thus the decline in heat dissipation capacity cannot be detected in a timely manner. However, in this embodiment of the invention, the driving relationship between the operating current and winding temperature is predetermined by a thermal state constraint model. Even if the statistical correlation between the two is very low during this period, the cloud data processing platform will still calculate the winding temperature based on the power loss corresponding to the current operating current and the nominal equivalent thermal resistance, according to this driving relationship. This allows the platform to detect the phenomenon where the actual temperature is consistently higher than the calculated temperature even before the winding temperature exceeds its limit.

[0084] The above distinction method does not depend on the specific type of the monitored object. For different types of monitored objects, it is only necessary to call the physical constraint model that matches them.

[0085] Furthermore, the physical evolution relationship includes a state change mapping relationship and a time delay relationship. The state change mapping relationship describes how the state of the driving index in the current period or the previous period affects the state of the response index in the next period, and in this embodiment of the invention, it is given by the discrete form of the above-mentioned physical constraint model; the time delay relationship describes the time process during which the physical influence of the driving index is transmitted to the response index.

[0086] Taking the thermal state constraint model as an example, its thermal time constant is equal to the product of the motor's equivalent thermal capacity and equivalent thermal resistance, used to characterize the speed of the thermal response process. It should be noted that the thermal time constant does not mean that the winding temperature changes suddenly only after a certain time. Therefore, in practical applications, the time delay parameter can be determined based on the thermal time constant and a preset response ratio. For example, if the time required for the winding temperature to reach the preset ratio of its final change is used as the time delay parameter, then this time delay parameter can be expressed as:

[0087] ;

[0088] In the formula, This represents the time delay parameter corresponding to the physical evolution relationship; This represents the time coefficient corresponding to the preset response ratio; Indicates the equivalent heat capacity of the motor; This represents the equivalent thermal resistance of the motor. Optionally, the preset response ratio can range from 50% to 95%, and correspondingly, the time coefficient... The value of can range from 0.7 to 3.0. The above thermal equilibrium relationship is a first-order inertial element. The time required for its step response to reach the preset response ratio of the final change is equal to the product of the negative natural logarithm of the preset response ratio and the time constant. Therefore, when the preset response ratio is 50%, 63.2%, 90%, and 95%, the time coefficient is approximately 0.69, 1.0, 2.3, and 3.0, respectively. If the preset response ratio is lower than 50%...

[0089] In this embodiment of the invention, the aforementioned time delay relationship is used not only to describe the transmission process of physical influence, but also to determine the start time of the preset time range of the historical dynamic index data in S106. That is, the start time is determined based on the time delay relationship, and its specific determination method will be described in detail in S106.

[0090] Thus, the use of the physical constraint model provides a physical basis for the differentiation of indicators, so that the direction of action between indicators no longer depends on statistical correlation; the result of indicator differentiation determines the calculation object and the deviation comparison object of the predicted value of the response indicator in the subsequent steps; and the physical evolution relationship provides the specific form and time span of the response indicator derived from the driving indicator.

[0091] S103. Calculate the current predicted value of the response indicator based on the physical evolution relationship, driving indicator data, and response indicator data, and calculate the deviation between this predicted value and the currently acquired response indicator data to obtain the degree of physical consistency deviation. The period to which the driving indicator data belongs is determined by the time delay relationship corresponding to the physical evolution relationship: if the time delay is less than the processing period, the driving indicator data of the current period is used; otherwise, the driving indicator data of the previous period is used. When the physical evolution relationship is a cross-period recursive form, the response indicator data is taken from the response indicator data of the previous period.

[0092] The formula for calculating the degree of deviation from physical consistency is:

[0093] ;

[0094] in, Indicates the i-th response indicator at The degree of physical consistency deviation at any given moment; Indicates the i-th response indicator at The response indicator data is obtained from the actual data acquisition equipment at all times; Indicates the i-th response indicator at The predicted value of the response index obtained from the physical constraint model at any time; This represents the permissible deviation scale of the i-th response indicator under normal operating conditions, that is, the maximum allowable deviation between the actual collected value and the predicted value of the response indicator under normal operating conditions.

[0095] It should be understood that the driving index data actually collected by the data acquisition equipment refers to the situation described in S102 where the same value has different physical roles in different physical evolution relationships. Specifically, it refers to the case where the same value has different physical roles in different physical evolution relationships. Each response metric in The value is the actual value collected by the data acquisition equipment at any given moment. This value serves two purposes: firstly, it acts as the actual value of the response index within the physical evolution of the current cycle, and its deviation is calculated compared to the predicted value of the response index derived from the driving index data of the previous cycle; secondly, it also participates in the calculation as driving index data in the state change mapping relationship of the next cycle, and in the path segment downstream of the response index. Therefore, the deviation calculation expressed in the above formula is performed between the actual collected value and the predicted value of the same response index at the same moment.

[0096] It should be understood that the predicted value of the response index is derived from the driving index data and the response index state of the previous period based on the state change mapping relationship determined by the physical constraint model. It is used to characterize the state that the response index should have reached in the current period under the condition that no abnormality has occurred in the monitored object.

[0097] Specifically, taking winding temperature as an example, by discretizing the above thermal state constraint model, the predicted winding temperature for the current cycle can be obtained:

[0098] ;

[0099] In the formula, This indicates the predicted winding temperature for the current cycle. This indicates the actual winding temperature collected in the previous cycle; This indicates the power loss in the previous cycle; This indicates the ambient temperature of the previous period. Indicates the equivalent heat capacity of the motor; This indicates the equivalent thermal resistance of the motor.

[0100] When copper loss is used as the main loss, the power loss of the previous cycle is taken as the product of the square of the operating current of the previous cycle and the equivalent resistance of the winding. Then the predicted value of the winding temperature can be expressed as:

[0101] ;

[0102] Similarly, for the speed and vibration amplitude response indicators, discretizing the above motion state constraint model can yield the predicted value for the current cycle. This calculation process does not use the actual speed data collected in the current cycle.

[0103] It should be noted that the above-mentioned predicted values ​​of response indicators are only used to evaluate whether the actual collected dynamic indicator data conforms to the normal physical evolution law, and are not used to replace the actual collected values. Therefore, the actual state recorded in the dynamic situation results will not be changed due to the prediction error of the physical constraint model itself.

[0104] By calculating the deviation between the data acquired in the current period and the predicted value of the response indicator using the formula described above, the degree of physical consistency deviation of the response indicator can be obtained. Optionally, It can be determined based on historical data statistics of the monitored object under normal operating conditions, or it can be determined based on equipment design parameters; in one implementation method, The value can be 2 to 4 times the standard deviation of the predicted residual under normal operating conditions, or it can be 2% to 10% of the rated value of the response index. In the absence of historical normal operating data, an empirical value can be taken as a percentage of the rated value. For example, for winding temperature, The value range can be from 3℃ to 10℃, preferably from 4℃ to 8℃; for rotational speed, The value can range from 1% to 5% of the rated speed. This invention is not limited to this.

[0105] For example, the actual winding temperature collected in the current cycle is 82℃, while the predicted winding temperature calculated based on the thermal state constraint model is 79℃. The allowable deviation scale for the winding temperature is... The temperature is 5℃, and in the formula, The permissible deviation scale for winding temperature under normal operating conditions is represented by a physical consistency deviation of 0.6. Since a larger physical consistency deviation indicates a greater degree of deviation from normal physical evolution, this deviation can be used to identify state changes that have not yet crossed the safety boundary but no longer conform to normal physical evolution.

[0106] In the application scenario of dust accumulation in cooling fans, since the actual equivalent thermal resistance of the motor is already greater than the nominal equivalent thermal resistance recorded in the physical constraint model, the winding temperature prediction value calculated according to the above discrete form will be continuously lower than the actual collected winding temperature. The deviation between the two accumulates cycle by cycle, causing the physical consistency deviation to gradually increase from 0.7 to 1.4, so that it is identified when the winding temperature is still tens of degrees Celsius away from the safety upper limit. During this process, the winding temperature is always within the safe range and the single-cycle change rate is very low, so the judgment based on the safety threshold and the change rate will not be triggered.

[0107] Therefore, the state change mapping relationship provides a specific form for extrapolating the current cycle's response index from the previous cycle's driving index, making the predicted response index a unique reference value that depends only on physical laws and historical inputs, and not on the actual value of the current response index. The deviation between this independent reference value and the currently acquired data is calculated, and the deviation is scaled according to an allowable deviation measure. Normalization allows for comparison of the degree of deviation of response indicators with different dimensions and orders of magnitude on the same scale, and the setting of thresholds can be standardized. The three factors work together to achieve the technical effect of detecting deviations from normal physical evolution before dynamic indicators cross the safety boundary, and to provide a third judgment dimension for the steady state judgment in S104 that is independent of the safety boundary and the rate of change.

[0108] S104. Calculate the distance between the multidimensional dynamic index data and the preset safety boundary, as well as the rate of change of the dynamic index, and determine whether the stability condition is met by combining the degree of deviation from physical consistency. The stability condition includes the degree of deviation from physical consistency being within the allowable error range of the corresponding physical constraint model, the distance between the multidimensional dynamic index data and the preset safety boundary being greater than the preset safety margin threshold, and the rate of change of the dynamic index being less than the maximum allowable rate of change of the corresponding index.

[0109] Specifically, if the The safe interval corresponding to each dynamic indicator is: In the formula, This indicates the lower limit of the safety of this dynamic indicator; If the dynamic indicator represents the upper limit of safety, then the distance between the dynamic indicator and the nearest safety boundary is:

[0110] ;

[0111] In the formula, Indicates the first The distance between a dynamic indicator and the nearest safety boundary in the current period; Indicates the first The distance is the actual value of a dynamic indicator in the current period. Optionally, if the dynamic indicator only has a safety upper limit, the distance is the difference between the safety upper limit and the current value; if the dynamic indicator only has a safety lower limit, the distance is the difference between the current value and the safety lower limit. Further, to eliminate the influence of different dynamic indicator dimensions, the distance can be divided by the width of the safety interval to obtain the normalized safety boundary distance. When using the normalized safety boundary distance, the corresponding preset safety margin threshold should also adopt the same normalized form to ensure comparability. For example, if the safe range for winding temperature is 20℃ to 120℃, its safe range width is 100℃, and its safety upper limit is 120℃, and the actual winding temperature collected in the current period is 82℃, then the distance between the current winding temperature and the safety upper limit is 38℃.

[0112] The rate of change of this dynamic indicator can be calculated based on the dynamic indicator data of the current period and the previous period:

[0113] ;

[0114] In the formula, Indicates the first The rate of change of a dynamic indicator in the current period; Indicates the first The actual value of each dynamic indicator in the previous period. Optionally, to reduce the impact of disturbances at individual sampling points, the value can also be taken as the sum of the values ​​of the current period and the previous period. The absolute value of the difference between the dynamic indicator values ​​of each period, divided by The average rate of change is obtained by multiplying the product of the processing cycle length and the average rate of change. This indicates the number of cycles involved in the calculation of the rate of change, and its value can range from 2 to 20.

[0115] In this embodiment of the invention, the output module makes judgments according to the following three conditions. For the first... The physical consistency condition for each response metric is:

[0116] ;

[0117] In the formula, This indicates the degree of physical consistency deviation of the i-th response metric in the current period. Indicates the first The permissible error threshold for the physical constraint model corresponding to each response index. Optionally, the permissible error threshold can range from 0.5 to 3.0, since the degree of deviation from physical consistency has been scaled according to the permissible deviation scale. After normalization, the allowable error threshold of 1 indicates that the deviation between the actual value and the predicted value exactly reaches the allowable deviation scale. Therefore, the threshold should be centered around 1. For response indicators that directly correspond to the safety boundary of the monitored object, the allowable error threshold can be taken as a smaller value within the above range to improve the judgment sensitivity of the response indicator.

[0118] No. The safety margin conditions corresponding to each dynamic indicator are:

[0119] ;

[0120] In the formula, This represents the distance between the i-th dynamic indicator and the nearest safety boundary in the current period. Indicates the first Each dynamic indicator corresponds to a preset safety margin threshold. Optionally, the preset safety margin threshold can range from 5% to 20% of the corresponding safety interval width. The preset safety margin threshold is not less than the product of the maximum allowable rate of change of the dynamic indicator and the preset response time. The preset response time refers to the time required from the switch from the processing mode to the full processing mode until the intervention of the maintenance personnel or protection device, and its value range can be from 1 minute to 30 minutes. For example, for winding temperature, its maximum allowable rate of change is 2℃ / min, and the preset response time is 5 minutes, then the preset safety margin threshold should not be less than 10℃. At the same time, the preset safety margin threshold should not exceed 20% of the safety interval width; otherwise, load fluctuations under normal operating conditions will cause this condition to be frequently triggered, thus causing the monitored object to remain in the full processing mode for a long time and lose the effect of reducing resource consumption. Accordingly, for winding temperature, the preset safety margin threshold can range from 10℃ to 20℃.

[0121] No. The rate of change conditions corresponding to each dynamic indicator are:

[0122] ;

[0123] In the formula, This indicates the rate of change of the i-th dynamic indicator in the current period. Indicates the first The maximum permissible rate of change corresponds to each dynamic indicator. Optionally, this maximum permissible rate of change can be determined as 1.2 to 3 times the maximum rate of change of the dynamic indicator during the process of increasing from no-load to full-load under normal operating conditions. For example, in this embodiment of the invention, when the motor increases from no-load to full-load, the maximum temperature rise rate of the winding is approximately 1℃ / min. Taking twice this, i.e., 2℃ / min, as the maximum permissible rate of change, the range of the maximum permissible rate of change for the winding temperature can be 1.2℃ / min to 3℃ / min. It should be noted that the above-mentioned rate of change and the maximum permissible rate of change are both based on the processing cycle length. As a time reference, a uniform time unit should be used in the same embodiment.

[0124] When all response indicators involved in the stability assessment meet the physical consistency condition, all dynamic indicators involved in the safety assessment meet the safety margin condition, and all dynamic indicators involved in the rate of change assessment meet the rate of change condition, the monitored object is determined to meet the stability condition, and the stability state flag for the current period can be set to 1. If any of the above three conditions is not met by any dynamic indicator, the monitored object is determined to not meet the stability condition, and the stability state flag for the current period is set to 0. The indicators involved in the stability assessment are all dynamic indicators involved in the invoked physical constraint model or a pre-configured set of key response indicators.

[0125] In another implementation, when there are a large number of response indicators involved in the judgment, the physical consistency deviation of each response indicator can be weighted and summed to obtain a comprehensive physical consistency deviation, and this comprehensive physical consistency deviation can be compared with the corresponding comprehensive allowable error threshold. This embodiment of the invention is not limited thereto.

[0126] It should be noted that the above three conditions describe the state of the monitored object from different perspectives: the degree of deviation from physical consistency is used to characterize whether the current state conforms to the normal physical evolution law, the safety boundary distance is used to characterize the margin of the current state from the unsafe interval, and the rate of change of dynamic indicators is used to characterize how fast the current state evolves.

[0127] Alternatively, in another implementation, only in continuous The stable state flag for each processing cycle is 1, meaning continuous. The sum of the stable state flags for each processing cycle equals Only when this condition is met is a switch from full processing mode to low-resource processing mode allowed, where, This represents the preset number of consecutive stable periods, which can range from 3 to 30. The product of this number of consecutive stable periods and the processing cycle length is the latency required to switch from full processing mode to low-resource processing mode. An excessively long latency will shorten the actual running time of low-resource processing mode, thus weakening the effect of reducing resource consumption. This avoids frequent switching of processing modes due to occasional data fluctuations in a single cycle. It should be noted that the consecutive period condition does not need to be met when switching from unstable mode to full processing mode; that is, the switching of processing modes is asymmetrical in both directions to ensure that full data processing capability can be immediately restored in the event of an anomaly. It should also be noted that this consecutive stable period condition is an optional additional constraint and does not constitute a limitation on the aforementioned stability condition judgment itself.

[0128] The above three conditions work together to ensure that both slow-deterioration and abrupt anomalies can be identified. The timing of switching processing modes will not result in the loss of data on the anomaly formation process due to overly slow judgment, nor will it lead to frequent mode switching due to overly sensitive judgment.

[0129] S105. If satisfied, a low-resource processing mode is applied to the multidimensional dynamic indicator data to generate dynamic situational results. The low-resource processing mode includes reducing the upload frequency of multidimensional dynamic indicator data, reducing the cloud data storage frequency, and converting the multidimensional dynamic indicator data within the statistical period into summary data. The summary data includes the maximum value, minimum value, average value, and trend parameters. When generating summary data using the low-resource processing mode, key response indicators in the response indicator data are determined based on the physical evolution relationship. Complete summary data is generated for the key response indicators, and basic summary data containing only the average value is generated for the remaining dynamic indicators besides the key response indicators.

[0130] For example, in full processing mode, edge processing nodes upload raw multidimensional dynamic indicator data to the cloud data processing platform once per second. After entering low-resource processing mode, edge processing nodes still collect raw dynamic indicator data once per second, but generate and upload summary data every 60 seconds. Optionally, the statistical period can range from 10 seconds to 600 seconds. This statistical period should be longer than the shortest normal fluctuation period of the key response indicator, and should ensure that at least three to five sets of summary data can be generated within a time span corresponding to a time delay. That is, the statistical period should not exceed one-third of the total time delay of the physical evolution path corresponding to the key response indicator, thus providing a basis for extrapolating the trend parameters of the early switching mechanism described later.

[0131] The data upload frequency in low-resource processing mode can be reduced by a factor of 10 to 120 compared to the data upload frequency in full processing mode. The upper limit is constrained by the length of the statistical period, while the lower limit should ensure a sufficiently significant reduction in resource consumption. The reduction factor for the frequency of raw cloud data storage can be the same as the reduction factor for the upload frequency. In implementations requiring further reduction in cloud storage overhead, uploaded data can be stored at preset sampling intervals, further reducing the cloud storage frequency to one-half to one-fifth of the upload frequency.

[0132] It should be noted that in low-resource processing mode, the periodic calculations of physical consistency deviation, security boundary distance, and dynamic indicator change rate are performed by edge processing nodes, while the cloud data processing platform generates dynamic situational results based on summary data. This ensures that the granularity of stability condition judgment does not decrease due to the switching of processing modes, while also substantially reducing the amount of data received, stored, and computed in the cloud.

[0133] Let a statistical period include There are sampling points, where, This represents the number of sampling points within a statistical period, and its value is equal to the ratio of the length of the statistical period to the local sampling period. The number of sampling points within this statistical period... The maximum value of each dynamic indicator is taken as the largest among the indicator values ​​of all sampling points within the statistical period, the minimum value is taken as the smallest among the indicator values ​​of all sampling points, and the average value is taken as the arithmetic mean of the indicator values ​​of all sampling points; the trend parameter of the dynamic indicator can be expressed as:

[0134] ;

[0135] In the formula, Indicates the first The changing trend parameter of a dynamic indicator within a statistical period; Indicates the start time of the statistical period The index value; Indicates the end time of the statistical period The indicator value is defined as follows: a value greater than zero indicates an overall upward trend for the dynamic indicator within the statistical period, while a value less than zero indicates an overall downward trend. It should be noted that although this trend parameter and the rate of change of the dynamic indicator have the same dimensions, their meanings are different: the trend parameter is a signed quantity, and its statistical interval is one statistical period, used to characterize the unidirectional tendency of the dynamic indicator over a longer time scale; while the rate of change of the dynamic indicator is a non-negative quantity, and its statistical interval is one processing period, used to characterize the speed of change of the dynamic indicator between adjacent periods.

[0136] In this embodiment of the invention, key response indicators can be determined based on indicator attributes pre-recorded in the physical constraint model. Optionally, a response indicator that meets at least one of the following conditions can be determined as a key response indicator: the response indicator directly corresponds to the safety boundary of the monitored object; the response indicator serves as the terminal state quantity of the physical evolution relationship; the response indicator participates in the switching judgment between low-resource processing mode and full processing mode; the response indicator is used to trigger the protection action or alarm action of the monitored object. Optionally, the proportion of key response indicators to the total number of dynamic indicators can be from 10% to 50%.

[0137] Specifically, in this embodiment of the invention, winding temperature and vibration amplitude are determined as key response indicators, and complete summary data including maximum value, minimum value, average value and trend parameters are generated for the winding temperature and vibration amplitude respectively; while for other dynamic indicators other than key response indicators such as ambient temperature and operating voltage, only basic summary data including average value is generated.

[0138] For example, in a specific application scenario, a motor device collects seven dynamic indicators. In full processing mode, the upload frequency is once per second, resulting in approximately 600,000 data entries uploaded per device per day. After entering low-resource processing mode and setting the statistical period to 60 seconds, 1440 statistical periods are generated daily. In each statistical period, four summary values ​​are generated for the two key response indicators, and one summary value is generated for each of the remaining dynamic indicators (excluding the five key response indicators). This results in approximately 19,000 data entries uploaded per device per day, a reduction of about 97% compared to the full processing mode. If full summary data is generated for all seven dynamic indicators without distinguishing between key response indicators, the number of data entries uploaded per day is approximately 40,000. Therefore, by adopting a differentiated summary method, the data volume is further reduced compared to the low-resource processing mode, while the peak values ​​and variation trends of winding temperature and vibration amplitude within the statistical period are fully preserved.

[0139] S106. If not satisfied, the full processing mode is adopted and a dynamic situation result is generated; wherein, if the current process is switching from low-resource processing mode to full processing mode, the historical dynamic indicator data within the preset time range before the anomaly occurred is called and fused with the current multi-dimensional dynamic indicator data to generate a dynamic situation result.

[0140] Specifically, the complete processing mode may include: uploading raw data of multi-dimensional dynamic indicators at a full upload frequency; storing raw dynamic indicator data in the cloud at a full storage frequency; increasing the update frequency of dynamic situation results; and performing physical consistency analysis, security boundary analysis, and rate of change analysis on all multi-dimensional dynamic indicators. Optionally, the dynamic situation results generated under the complete processing mode may at least include the current actual dynamic indicator value, the predicted value of the response indicator, the degree of physical consistency deviation, the security boundary distance, the rate of change of dynamic indicators, dynamic indicators that do not meet the stability conditions, the mode switching time, and the dynamic indicator change process before and after the anomaly trigger.

[0141] In this embodiment of the invention, when the current processing mode switches from low-resource processing mode to full processing mode, the exception trigger time is first determined. Optionally, the anomaly trigger time can be any of the following moments: the moment when the physical consistency deviation first exceeds the allowable error range; the moment when the distance between the dynamic indicator and the safety boundary first does not exceed the preset safety margin threshold; the moment when the rate of change of the dynamic indicator first reaches or exceeds the maximum allowable rate of change; and the moment when the predicted distance, as described below, first falls below the preset predicted safety margin threshold. Each initial exceedance moment is determined in real time by the edge processing node during the cycle-by-cycle calculation of the physical consistency deviation, safety boundary distance, and rate of change of the dynamic indicator in low-resource processing mode, and the timestamp of that moment is recorded as the anomaly trigger time when the corresponding conditions are met. When switching to full processing mode, this timestamp Along with historical dynamic indicator data prior to the anomaly, the edge processing nodes upload the data to the cloud data processing platform. Thus, the anomaly trigger time is determined by the edge processing nodes based on local raw data. Even if the cloud only receives summary data at the statistical period granularity in low-resource processing mode, it can still accurately determine the anomaly trigger time and the start time of historical backtracking without requiring the cloud to continuously receive high-frequency raw data.

[0142] Furthermore, based on the response indicator that triggers the switching of processing modes, the physical evolution path corresponding to that response indicator is determined. Let this response indicator be... The corresponding physical evolution path is In the formula, Represents a physical evolution path; This indicates the driving indicator at the beginning of the path; This indicates an intermediate state indicator that follows the driving indicator in the path; This indicates the response metrics at the end of the path; This indicates the number of dynamic indicators included in the path. The path contains the [number of] [dynamic indicators]. The time delay parameter between adjacent indicators is denoted as Then, when each physical process is passed sequentially, the total time delay corresponding to this physical evolution path is: ;

[0143] In the formula, Representing the physical evolution path The corresponding total time delay; This indicates the first [item] in the path. The time delay parameter between adjacent indicators.

[0144] Optionally, if the same response indicator corresponds to multiple physical evolution paths, the largest total time delay among all physical evolution paths is taken as the historical backtracking duration, thereby ensuring that the historical dynamic indicator data can cover all driving processes that may lead to changes in the current state. The start time of the preset time range of this historical dynamic indicator data is equal to the anomaly trigger time minus the historical backtracking duration, and the preset time range is the time interval from the start time to the anomaly trigger time. Optionally, to balance data integrity and call overhead, a backtracking margin coefficient can be multiplied by the historical backtracking duration, and the preferred value range of the backtracking margin coefficient is 1.1 to 1.5. The time delay parameters of each path segment in the physical evolution path are either calculated from equipment design parameters or calibrated from normal operation data, both of which have a deviation of 10% to 30%. Therefore, a backtracking margin coefficient of 1.1 to 1.5 can cover the above deviation.

[0145] For example, the physical evolution path corresponding to abnormal winding temperature is that the operating current is transferred to the winding temperature through power loss, i.e. The time delay from a change in operating current to a corresponding change in power loss is 5 seconds, and the time delay from a change in power loss to a corresponding change in winding temperature is 300 seconds. Therefore, the total time delay for this physical evolution path is 305 seconds. If the abnormal trigger time is 12:00:00, the starting time for the historical dynamic indicator data is 11:54:55, meaning that the original multidimensional dynamic indicator data within the range of 11:54:55 to 12:00:00 will be retrieved.

[0146] Specifically, edge processing nodes can extract raw dynamic indicator data within the aforementioned time range from the circular buffer and send it to the cloud data processing platform. Optionally, the capacity of the circular buffer should be no less than the product of the largest of the total time delays of the physical evolution paths corresponding to each response indicator of the monitored object, the local sampling frequency, the number of dynamic indicators, and the storage space occupied by a single data point, with a margin of 1.5 to 3 times reserved on this basis. The cloud data processing platform fuses this historical dynamic indicator data with the current multidimensional dynamic indicator data collected after switching to the full processing mode according to timestamps to obtain a continuous time series covering the period before and after the anomaly formation, and generates dynamic situational results based on this continuous time series.

[0147] It should be noted that since the scope of historical dynamic indicator data is determined based on the time delay relationship corresponding to the physical evolution path, rather than based on a fixed duration, for physical processes with long time delays such as heat transfer, it can trace back to a time range sufficient to cover its driving process; while for physical processes with short time delays such as electrical processes, it will not call up a large amount of irrelevant data due to an excessively large tracing range, thus ensuring the integrity of the anomaly formation process while avoiding the waste of historical data retrieval.

[0148] For example, taking motor equipment as an example, if a fixed backtracking duration is used, the backtracking duration needs to be uniformly set to more than 300 seconds to cover the heat transfer process. However, when an anomaly is triggered by the power state constraint model, the total time delay of its physical evolution path is usually less than 1 second. In this case, a 300-second backtracking duration would increase the amount of data called by more than two orders of magnitude, and most of this data would be unrelated to the formation process of the anomaly. Using the method of the embodiment of the present invention, the thermal state anomaly backtracks for 305 seconds and the power state anomaly backtracks for 1 second, both of which exactly cover their respective driving processes.

[0149] It can be seen that the time delay relationship determined in S102 and the historical data retrieval in this step work together to produce the technical effect of neither omitting the driving process nor redundantly calling historical data. This makes up for the lack of fine-grained historical data in the cloud under the low-resource processing mode, so that the resource savings brought about by the low-resource processing mode do not come at the cost of a decrease in the ability to trace anomalies.

[0150] Among these measures, when processing multidimensional dynamic indicator data using a low-resource processing mode, an advance switching mechanism is also included, specifically:

[0151] Based on the physical evolution relationship and the driving index data of the current cycle, the predicted value of the response index for each prediction cycle within the future preset time range is calculated recursively, and the predicted value of the response index at the end of the future preset time range is determined as the expected value of the response index.

[0152] Calculate the predicted distance between the expected value of the response index and the preset safety boundary, and determine whether the predicted distance is less than the preset predicted safety margin threshold;

[0153] If so, switch the multidimensional dynamic indicator data from low-resource processing mode to full processing mode.

[0154] Specifically, during the low-resource processing mode, edge processing nodes continue to acquire the driving indicator data for the current period according to their local sampling cycle. Let this future preset time range include... For each prediction period, in the formula, This indicates the number of prediction periods. Optionally, the length of this future preset time range can be 0.5 to 3 times the total time delay of the physical evolution path corresponding to the response index; correspondingly, the number of prediction periods... This is equal to the ratio of the length of the preset future time range to the length of the processing cycle, and its value can range from 3 to 120. For example, in this embodiment of the invention, the total time delay corresponding to the winding temperature is 305 seconds, then the preset future time range can be 300 seconds to 600 seconds. When the processing cycle length is 60 seconds, Accordingly, a value of 5 to 10 is chosen. For the winding temperature, it can be calculated recursively cycle by cycle according to the thermal state constraint model:

[0155] ;

[0156] In the formula, This indicates the prediction period number, with values ​​ranging from 1 to... ; Indicates the first Predicted winding temperature for each prediction cycle; Indicates the first Predicted power loss for each prediction period; Indicates the first The predicted ambient temperature for each prediction cycle. The initial value for the above recursion is the actual winding temperature collected in the current cycle.

[0157] Optionally, in one implementation, the driving indicators for each future forecast period remain unchanged based on the actual values ​​of the current period; in another implementation, the driving indicators can be extrapolated based on the trend parameters already obtained under low-resource processing mode, for example, the predicted operating current values ​​for each future forecast period can be expressed as:

[0158] ;

[0159] In the formula, Indicates the first Predicted operating current value for each prediction cycle; This parameter represents the trend of the operating current within the statistical period. Furthermore, the predicted power loss for each future prediction period can be taken as the product of the square of the predicted operating current for the corresponding prediction period and the equivalent resistance of the winding. Optionally, for drive indicators that change slowly, such as ambient temperature, the actual value of the current period can be kept constant; for drive indicators that change rapidly, such as operating current, extrapolation can be performed in the manner described above. However, this embodiment of the invention is not limited to this.

[0160] The end of the future preset time range, i.e. the first The predicted values ​​of the response indicators for each forecast period are determined as the expected values ​​of the response indicators.

[0161] It should be noted that the above prediction results serve only as triggering conditions for improving the granularity of data processing, and are not directly used to determine that the monitored object has experienced anomalies, nor are they written into the dynamic situation results as anomaly conclusions. Therefore, even if the prediction results contain some errors, they will not replace the actual collected data to form erroneous anomaly judgments. Their role is to enable the system to restore full data processing capabilities in advance during the formation stage of potential risks, rather than after dynamic indicators have actually crossed the limits, thereby ensuring that the original dynamic indicator data during the anomaly formation stage is completely uploaded and stored.

[0162] It is evident that the three technical features of trend parameters, physical evolution relationships, and prediction distance judgment under the low-resource processing mode work together to produce the technical effect that the low-resource processing mode itself has the judgment ability required to exit the mode. This resolves the contradiction between reducing data processing granularity and maintaining anomaly detection capability, allowing the system to restore the complete processing mode during the risk formation stage without relying on continuous high-frequency data uploads.

[0163] When the deviation of physical consistency exceeds the allowable error range of the corresponding physical constraint model, the physical evolution path from the driving index to the response index through one or more intermediate state indices is determined according to the physical evolution relationship. The physical evolution path is divided into multiple path segments composed of two adjacent dynamic indices. The physical consistency deviation corresponding to each path segment and the ratio of the physical consistency deviation to the allowable error threshold corresponding to the path segment are calculated to obtain the relative exceedance degree of each path segment. The physical transmission link corresponding to the path segment with the largest relative exceedance degree and greater than 1 is determined as the anomaly occurrence location, and a dynamic situation result containing the anomaly location is generated.

[0164] It should be understood that this physical evolution path, determined by the physical constraint model, refers to a directed sequence of indicators that progresses from the driving indicator through one or more intermediate state indicators to the response indicator, rather than a general correlation temporarily established based on statistical correlation. For example, the physical evolution path corresponding to the thermal state of a motor is... Wherein, the operating current is the driving index, the power loss is the intermediate state index, and the winding temperature is the response index; the physical evolution path corresponding to the motor's motion state can be... In this embodiment of the invention, the portion of the physical evolution path consisting of two adjacent dynamic indices is called a path segment.

[0165] When the same response indicator corresponds to multiple physical evolution paths, the relative exceedance degree of each path segment in each physical evolution path is calculated first. Then, the location of the anomaly is determined in all path segments in the following manner: First, all path segments with a relative exceedance degree greater than 1 are identified as candidate anomaly links, and they are sorted in descending order of relative exceedance degree and written into the dynamic situation results, instead of only taking the one with the largest relative exceedance degree, to avoid missed reporting when multiple paths are abnormal at the same time; Second, when there are multiple candidate anomaly links and the difference in their relative exceedance degree is within the preset tolerance range, they are given as parallel candidate anomaly links; Third, when the same dynamic indicator is shared by multiple physical evolution paths, and the adjacent path segments where the shared indicator is located exceed the limit in multiple paths, the anomaly is attributed to the physical transmission link corresponding to the shared indicator, instead of repeatedly attributing it to each path; Fourth, the investigation priority is determined in descending order of relative exceedance degree, from near to far from the corresponding security boundary, and whether it is located on the path where the critical response indicator is located. For convergent structures formed when a response indicator is simultaneously influenced by multiple driving indicators, corresponding driving-to-response path segments are constructed based on the contribution of each driving indicator to the predicted value of the response indicator, and attribution is performed according to the degree of physical consistency deviation of the contributions of each driving indicator.

[0166] Specifically, the anomaly localization module establishes local physical mapping relationships for two adjacent dynamic indicators in the physical evolution path and calculates the degree of physical consistency deviation between adjacent indicators. It should be noted that the above calculation by path segment is based on the premise that intermediate state indicators in the physical evolution path are observable. For example, power loss can be calculated by subtracting mechanical output power from input power. When an intermediate state indicator cannot be directly collected or calculated from other collected indicators, it is incorporated into an adjacent observable indicator to form an equivalent path segment, ensuring that the degree of physical consistency deviation of each path segment can be calculated. (Still using the physical evolution path...) For example.

[0167] For the path segment from operating current to power loss, the power loss predicted based on the operating current is the product of the square of the current cycle's operating current and the equivalent resistance of the winding; while the actual power loss is the input power of the current cycle minus the mechanical output power, where the mechanical output power is the product of the load torque and the angular velocity. The degree of physical consistency deviation of this path segment is:

[0168]

[0169] In the formula, This indicates the degree of physical consistency deviation of the path segment from operating current to power loss in the current cycle; This indicates the actual power loss during the current cycle; This represents the power loss predicted based on the operating current for the current cycle. It represents the allowable deviation of power loss under normal operating conditions, and its value ranges from 1% to 8% of the motor's rated power.

[0170] For the path segment from power loss to winding temperature, based on the actual power loss and winding temperature of the previous cycle, the winding temperature of the current cycle is predicted according to the discrete form of the thermal state constraint model. The degree of physical consistency deviation for this path segment is then:

[0171] ;

[0172] In the formula, This indicates the degree of physical consistency deviation of the path segment from power loss to winding temperature in the current cycle; This represents the actual winding temperature collected in the current cycle. From this, we can obtain the degree of physical consistency deviation corresponding to each segment of the physical evolution path.

[0173] It should be noted that in calculating the physical consistency deviation of each path segment, the period of the driving index data used is determined by the time delay parameter corresponding to that path segment: when the time delay parameter is less than the processing cycle length, the driving index data of the current cycle is used; when the time delay parameter is not less than the processing cycle length, the driving index data of the previous cycle is used. For example, the time delay of the path segment from operating current to power loss is 5 seconds, which is less than the processing cycle length of 60 seconds, so the operating current of the current cycle is used; the time delay of the path segment from power loss to winding temperature is 300 seconds, which is greater than the processing cycle length, so the power loss and winding temperature of the previous cycle are used.

[0174] Furthermore, when the allowable error thresholds for each path segment are the same, the physical consistency deviations of each path segment can be directly compared to determine the location of the anomaly. When the allowable error thresholds for each path segment are different, the physical consistency deviations of each path segment can be further normalized relative to its respective allowable error threshold to obtain the relative exceedance degree. This allows for comparison of path segments with different allowable error thresholds on the same scale. Based on this, the relative exceedance degree of each path segment can be calculated:

[0175] ;

[0176] In the formula, Indicates the first The relative degree of exceeding limits for each path segment in the current cycle; Indicates the first The degree of physical consistency deviation of each path segment in the current cycle; Indicates the first The permissible error threshold corresponding to each path segment. It should be noted that... The permissible error threshold corresponding to the aforementioned response indicators They belong to different levels: the latter is used to determine whether a response indicator as a whole meets the physical consistency condition, while the former is used to determine whether a physical transmission link in the associated path corresponding to the response indicator meets the physical consistency condition. The two can take different values. Optionally, the allowable error threshold for this path segment can range from 0.5 to 2.0.

[0177] The physical transmission link corresponding to the path segment with the highest relative exceedance degree (greater than 1) is identified as the location of the anomaly.

[0178]

[0179] In the formula, This indicates the sequence number of the path segment with the highest relative degree of exceeding the limit. This indicates the relative degree of exceeding the limit for the j-th path segment in the current period.

[0180] For example, if the physical consistency deviation of the path segment from operating current to power loss is 0.6, and its corresponding allowable error threshold is 1, then the relative exceedance of this path segment is 0.6. If the physical consistency deviation of the path segment from power loss to winding temperature is 1.8, and its corresponding allowable error threshold is 0.9, then the relative exceedance of this path segment is 2. Therefore, the location of the anomaly is determined to be the heat transfer link corresponding to the power loss to winding temperature, i.e., a decrease in heat dissipation capacity, rather than an abnormal current input.

[0181] For example, in another specific application scenario, the mechanical load driven by the motor may jam due to poor bearing lubrication, causing the load torque to increase, the operating current to rise, and the speed to decrease. This occurs along the physical evolution path... The calculations are as follows: For the path segment from operating current to electromagnetic torque, since the electromagnetic relationship between electromagnetic torque and operating current remains unchanged, its relative exceedance level is still below 1. However, for the path segment from electromagnetic torque to speed, since the actual decrease in speed exceeds the range calculated based on the moment of inertia and torque difference, its relative exceedance level is greater than 1. Therefore, the location of the anomaly is determined to be the mechanical transmission link corresponding to the electromagnetic torque to speed, thus pointing the investigation scope to the load side and transmission components. In contrast, if the above motor experiences a short circuit between turns in the stator winding, the relative exceedance level of the path segment from operating current to electromagnetic torque will be greater than 1, while the relative exceedance level of the path segment from electromagnetic torque to speed will still be below 1. In this case, the location of the anomaly is pointed to the electromagnetic conversion link. It can be seen that even when both the operating current and speed dynamic indicators are abnormal, the above method can still distinguish whether the anomaly occurs in the electromagnetic conversion link or the mechanical transmission link.

[0182] It should be noted that the above-mentioned location of the anomaly indicates the indicator transmission link where the physical consistency anomaly occurs, and is not directly equivalent to identifying the specific faulty component; however, since the location points to the specific physical transmission process, it can significantly narrow the scope of investigation and improve the interpretability of dynamic situation results compared to simply pointing out that a certain dynamic indicator has exceeded the limit.

[0183] Optionally, the dynamic situational results including the abnormal location can record the abnormal response indicators, the corresponding physical evolution path, the abnormal path segment, the degree of physical consistency deviation of each path segment, the relative degree of exceeding the limit of each path segment, and the dynamic indicator change process before and after the abnormality is triggered.

[0184] This demonstrates that even when multiple dynamic indicators are abnormal simultaneously, the physical transmission link of the abnormality can still be uniquely identified, thereby elevating the dynamic situation result from which indicator is abnormal to which physical link the abnormality occurred.

[0185] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0186] Therefore, the multi-dimensional dynamic situation indicator processing method of this invention provides a physical constraint model that not only distinguishes between driving and response indicators, but also provides a state change mapping relationship for calculating the predicted value of the response indicator, as well as a time delay relationship for determining the historical backtracking range and a physical evolution path for locating abnormal links. Since all judgments and processing are based on the same physical constraint model, data compression, early switching judgment, historical data backtracking, and abnormal link location under low-resource processing mode can operate collaboratively under the same set of physical criteria. This results in a significant reduction in resource consumption, while improving state observability, early anomaly detection, and result interpretability.

[0187] Example 2

[0188] like Figure 3 and Figure 4 As shown in the figure, this embodiment introduces a multi-dimensional dynamic situation indicator processing system, including:

[0189] The data acquisition module 201 is used to acquire multi-dimensional dynamic indicator data and object attributes corresponding to the monitored object;

[0190] The model invocation module 202 is used to invoke the physical constraint model matching the monitored object from the model database based on the object attributes, and to distinguish the driving indicator data and response indicator data with physical driving relationship and the physical evolution relationship between them from the multidimensional dynamic indicator data according to the physical constraint model.

[0191] The deviation calculation module 203 is used to calculate the current predicted value of the response index based on the physical evolution relationship, driving index data and response index data, and to calculate the degree of deviation of the physical consistency with the currently acquired response index data.

[0192] The judgment output module 204 is used to calculate the distance between the multi-dimensional dynamic index data and the preset safety boundary, as well as the rate of change of the dynamic index, and to judge whether the stability condition is met by combining the degree of deviation from physical consistency.

[0193] If the conditions are met, a low-resource processing mode is adopted for the multidimensional dynamic indicator data and dynamic situation results are generated.

[0194] If the conditions are not met, the full processing mode is adopted and a dynamic situational result is generated. If the current process is switching from the low-resource processing mode to the full processing mode, the historical dynamic indicator data within the preset time range before the anomaly occurred is called and fused with the current multi-dimensional dynamic indicator data to generate a dynamic situational result.

[0195] Specifically, the data acquisition module can be located at the edge processing node, used to acquire multi-dimensional dynamic indicator data through various data acquisition devices according to a preset local sampling period, and to obtain object attributes through device registration information, device configuration files, or a device management platform; the model calling module, the deviation calculation module, and the judgment output module can be located on the cloud data processing platform, or partially located at the edge processing node according to the actual computing power allocation. Optionally, in low-resource processing mode, the portion of the deviation calculation module and the judgment output module that performs cycle-by-cycle calculation is located at the edge processing node, but this embodiment of the invention is not limited thereto.

[0196] Optionally, the processing system may further include an anomaly localization module 205, used to determine the physical evolution path from the driving index to the response index through one or more intermediate state indices based on the physical evolution relationship when the degree of deviation from physical consistency exceeds the allowable error range of the corresponding physical constraint model. The physical evolution path is divided into multiple path segments composed of two adjacent dynamic indices. The degree of deviation from physical consistency corresponding to each path segment and the ratio of the degree of deviation from physical consistency to the allowable error threshold corresponding to the path segment are calculated respectively to obtain the relative exceedance degree of each path segment. The physical transmission link corresponding to the path segment with the largest relative exceedance degree and greater than 1 is determined as the location of the anomaly, and a dynamic situation result containing the anomaly location is generated.

[0197] It should be understood that the multi-dimensional dynamic situation indicator processing system according to embodiments of the present invention can correspond to the execution of the method in the embodiments of the present invention, and the above and other operations and / or functions of each module in the processing system are respectively for implementing Figure 1 For the sake of brevity, the corresponding processes of each method will not be elaborated here.

[0198] Therefore, the multi-dimensional dynamic situation indicator processing system of this invention distinguishes between driving indicator data and response indicator data through a physical constraint model and calculates the degree of physical consistency deviation. It combines the safety boundary distance and the rate of change of dynamic indicators to jointly determine the stability conditions and switches between low-resource processing mode and full processing mode accordingly. This can significantly reduce the data processing resource consumption while retaining key situation information, thereby greatly improving the accuracy and interpretability of dynamic situation results.

[0199] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for processing multi-dimensional dynamic situation indicators, characterized in that, Includes the following steps: Obtain multidimensional dynamic indicator data and object attributes corresponding to the monitored object; Based on object attributes, the physical constraint model matching the monitored object is retrieved from the model database. Based on the physical constraint model, the driving indicator data and response indicator data with physical driving relationship and the physical evolution relationship between the two are distinguished from the multidimensional dynamic indicator data. The predicted value of the current response index is calculated based on the physical evolution relationship, driving index data, and response index data, and the deviation from the currently acquired response index data is calculated to obtain the degree of physical consistency deviation. Calculate the distance between the multidimensional dynamic index data and the preset safety boundary, as well as the rate of change of the dynamic index, and combine the degree of deviation from physical consistency to determine whether the stability condition is met; If the conditions are met, a low-resource processing mode is adopted for the multidimensional dynamic indicator data and dynamic situation results are generated. If the conditions are not met, the full processing mode is adopted and a dynamic situational result is generated. If the current process is switching from the low-resource processing mode to the full processing mode, the historical dynamic indicator data within the preset time range before the anomaly occurred is called and fused with the current multi-dimensional dynamic indicator data to generate a dynamic situational result.

2. The method for processing multi-dimensional dynamic situation indicators according to claim 1, characterized in that, The physical constraint models include thermal state constraint models based on energy transfer relationships, motion state constraint models based on dynamic relationships, electrical state constraint models based on electrical characteristics, flow state constraint models based on fluid transport relationships, and remote sensing target state constraint models based on spatial variation relationships.

3. The method for processing multi-dimensional dynamic situation indicators according to claim 2, characterized in that, Physical evolution relationships include the time delay relationship in which driving index data affects response index data, as well as the mapping relationship of state changes; The start time of the preset time range for historical dynamic indicator data is determined based on the time delay relationship.

4. The method for processing multi-dimensional dynamic situation indicators according to claim 3, characterized in that, The formula for calculating the degree of deviation from physical consistency is: ； in, Indicates the i-th response indicator at The degree of physical consistency deviation at any given moment; Indicates the i-th response indicator at The response indicator data is obtained from the actual data acquisition equipment at all times; Indicates the i-th response indicator at The time interval represents the predicted value of the response index obtained based on the physical constraint model. This represents the allowable deviation scale for the i-th response index under normal operating conditions.

5. The method for processing multi-dimensional dynamic situation indicators according to claim 1, characterized in that, Stability conditions include the degree of deviation from physical consistency being within the allowable error range of the corresponding physical constraint model, the distance between the multidimensional dynamic index data and the preset safety boundary being greater than the preset safety margin threshold, and the rate of change of the dynamic index being less than the maximum allowable rate of change of the corresponding index.

6. The method for processing multi-dimensional dynamic situation indicators according to claim 5, characterized in that, The low-resource processing mode includes reducing the upload frequency of multidimensional dynamic indicator data, reducing the cloud data storage frequency, and converting multidimensional dynamic indicator data within the statistical period into summary data; the summary data includes the maximum value, minimum value, average value, and trend parameters.

7. The method for processing multi-dimensional dynamic situation indicators according to claim 6, characterized in that, When processing multidimensional dynamic indicator data using a low-resource processing mode, a pre-switch mechanism is also included, specifically: Based on the physical evolution relationship and the driving index data of the current cycle, the predicted value of the response index for each prediction cycle within the future preset time range is calculated recursively, and the predicted value of the response index at the end of the future preset time range is determined as the expected value of the response index. Calculate the predicted distance between the expected value of the response index and the preset safety boundary, and determine whether the predicted distance is less than the preset predicted safety margin threshold; If so, switch the multidimensional dynamic indicator data from low-resource processing mode to full processing mode.

8. The method for processing multi-dimensional dynamic situation indicators according to claim 7, characterized in that, When generating summary data using a low-resource processing mode, key response indicators in the response indicator data are determined based on the physical evolution relationship. Complete summary data is generated for the key response indicators, while basic summary data containing only average values ​​is generated for the remaining dynamic indicators other than the key response indicators.

9. The method for processing multi-dimensional dynamic situation indicators according to claim 1, characterized in that, When the deviation from physical consistency exceeds the allowable error range of the corresponding physical constraint model, the physical evolution path from the driving index to the response index is determined according to the physical evolution relationship. The index of the physical evolution path is used as a dynamic index. The physical evolution path is divided into multiple path segments composed of two adjacent dynamic indices. The deviation from physical consistency corresponding to each path segment and the ratio of the deviation from physical consistency to the allowable error threshold corresponding to the path segment are calculated to obtain the relative exceedance degree of each path segment. The physical transmission link corresponding to the path segment with the largest relative exceedance degree and greater than 1 is determined as the anomaly location, and a dynamic situational result containing the anomaly location is generated.

10. A system for processing multi-dimensional dynamic situation indicators, characterized in that, It includes: The data acquisition module is used to acquire multidimensional dynamic indicator data and object attributes corresponding to the monitored object; The model invocation module is used to retrieve physical constraint models that match the monitored objects from the model database based on object attributes. Based on the physical constraint models, it distinguishes driving indicator data and response indicator data with physical driving relationships and the physical evolution relationship between the two from multidimensional dynamic indicator data. The deviation calculation module is used to calculate the current predicted value of the response index based on the physical evolution relationship, driving index data and response index data, and to calculate the degree of deviation of the physical consistency with the currently acquired response index data. The output module is used to calculate the distance between the multi-dimensional dynamic indicator data and the preset safety boundary, as well as the rate of change of the dynamic indicators. It combines the degree of deviation from physical consistency to determine whether the stability condition is met, and generates dynamic situation results in the following manner: If the conditions are met, a low-resource processing mode is adopted for the multidimensional dynamic indicator data and dynamic situation results are generated. If the conditions are not met, the full processing mode is adopted and a dynamic situational result is generated. If the current process is switching from the low-resource processing mode to the full processing mode, the historical dynamic indicator data within the preset time range before the anomaly occurred is called and fused with the current multi-dimensional dynamic indicator data to generate a dynamic situational result.

Images