Elevator Digital Twin Virtual and Physical Sensor Data Calibration Method

Abstract

The present invention discloses a method for calibrating elevator digital twin virtual and physical sensor data, which includes: obtaining real-time data of multiple types of physical sensors in the elevator system under the reference environment and real-time data of virtual sensors in the elevator digital twin model at the same moment; preprocessing the data, comparing the collected physical sensor and virtual sensor data of the same category according to the distance function to obtain the error between the physical sensor and the virtual sensor at time t; comparing whether the error exceeds the set error threshold; if the error exceeds the threshold, the physical sensor data is corrected through linear regression calibration and offset correction functions, and the virtual sensor data at the next moment in the digital twin is updated. This application combines the measured data of physical sensors with the measured data of virtual sensors in the digital twin, calculates and calibrates the sensor deviation in real time, and can effectively solve the problem of single sensor error.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator digital twins, and particularly to a method for calibrating data of virtual and physical sensors of an elevator digital twin. Background Art

[0002] With the rapid development of China's economic construction, high-rise buildings have been built rapidly, and elevators have become an indispensable facility in people's daily production and life. At present, with the wide popularization of the Internet of Things, elevators have gradually applied the Internet of Things system. However, due to various uncertainties in the system itself, and the physical sensors of elevators also have instability during long-term operation. Because the hardware unit, communication unit of the physical sensor itself, as well as the surrounding environment of the sensor, may all cause the sensor to malfunction and generate various phenomena such as noise, offset, and drift. For example, there is a sensor with intermittent drift error in the elevator sensing system (traction machine temperature, vibration sensor). Due to the compact deployment space of the sensor and the heat radiation interference from the surrounding control cabinet, the temperature sensor will generate temperature and vibration deviations due to factors such as electromagnetic interference, heat radiation, and hardware abnormalities. Therefore, when facing a limited and unreliable sensing environment, it will have a negative impact on the data-driven applications in the elevator digital twin system. The digital twin technology integrates multi-physics, multi-disciplinary, and multi-scale attributes, and has the characteristics of ultra-realism, real-time synchronization, all elements, and uniqueness, and can realize the interactive integration of the physical world and the information world, so as to reflect the whole life cycle process of the corresponding physical equipment.

[0003] Currently, the retrieved digital twin virtual-real consistency method is the patent of "Determination and Interaction Method for Virtual-Reality Consistency of Complex System-Level Digital Twins" (application number 202211410916.5). This method discloses a method for determining and interacting the virtual-real consistency of complex system-level digital twins, describes the sub-spaces in the physical space and virtual space of complex systems, proposes a method for determining virtual-real consistency based on dynamic operation, and proposes specific operations and interaction methods between virtual and real sub-spaces;

[0004] This method establishes the operation and interaction process between the virtual space and the physical space according to the correlation relationship of the complex system digital twin body, but there are still some deficiencies:

[0005] 1. This technology has not considered how to ensure the accuracy and timeliness of data transmission between physical sensors and virtual sensors, and cannot further guarantee the effectiveness of the mutual feedback of digital twin virtual-real models.

[0006] 2. Traditional methods usually only focus on the error calibration of a single sensor, and have not considered the optimization of sensor calibration in the local or global environment of digital twins.

[0007] Therefore, a data calibration method is needed to ensure the reliability of physical sensors and virtual sensors, and to ensure the accuracy and safety of the elevator digital twin during its full life cycle operation. For this purpose, the present invention proposes a method for calibrating the data of virtual and physical sensors of an elevator digital twin. Summary of the Invention

[0008] Therefore, the purpose of the present invention is to provide a method and system for calibrating the data of virtual and physical sensors of an elevator digital twin, to solve the error problem existing between sensor data, and to ensure the consistency of virtual sensor and physical sensor data.

[0009] To achieve the above purpose, a method and system for calibrating the data of virtual and physical sensors of an elevator digital twin provided by the present invention includes the following steps:

[0010] S1. Obtain the real-time data of multiple types of physical sensors in the elevator system under the reference environment and the real-time data of virtual sensors in the elevator digital twin model at the same moment t;

[0011] S2. Preprocess the data, compare the collected data of the same type of physical sensors and virtual sensors according to the distance function, and obtain the error between the physical sensors and virtual sensors at moment t D fused (t);

[0012] S3. Compare the error D fused (t) to see if it exceeds the set error threshold ; if the error exceeds the threshold, correct the physical sensor data through linear regression calibration and offset correction function, and update the virtual sensor data at the next moment in the digital twin.

[0013] Further preferably, in S3, the distance function includes a direct distance function and an indirect distance function. The direct distance function is used to calculate the error between the physical sensor data and the virtual sensor prediction value, and the indirect distance function is used to calculate and verify the matching relationship between each sensor and calibrate the transfer error.

[0014] Further preferably, the direct distance function is expressed by the following formula:

[0015] ;

[0016] Wherein, is the error between the physical sensor and the virtual sensor calculated at moment ; is the physical sensor data, is the virtual sensor data.

[0017] Further preferably, the indirect distance function is expressed by the following formula:

[0018] ;

[0019] where: is the predicted value based on the virtual sensor for the physical sensor ; n is the number of physical sensors, m is the number of virtual sensors, is the error calculated between the predicted value and the physical sensor at time ; i is the index of the actual physical sensor dataset, and j represents the index of the corresponding virtual sensor dataset.

[0020] Further preferably, it further includes establishing a neural network to predict the data collected by various physical sensors according to the correlation between sensors, and obtaining

[0021]

[0022] where NN represents the neural network, f(Vj(t)) is the obtained predicted value; V 1 (t), V 2 (t), V m (t) are the input values.

[0023] Further preferably, in S3, when updating the virtual sensor data at the next moment in the digital twin, according to the error calculated by the indirect distance function, the virtual sensor model is adjusted using the weighted average method: ;

[0024] where: is the weight of the physical sensor, is the weight of the virtual sensor, and by optimizing the weighting coefficient, the error between the physical sensor and the virtual sensor is minimized, is the weighted average of the virtual sensor and the physical sensor.

[0025] Further preferably, in S2, when comparing according to the distance function, the distance function is the result of weighted fusion of the direct distance function and the indirect distance function, and is expressed by the following formula;

[0026] ;

[0027] where: is the fused error function; is the weight coefficient; is the direct distance function, is the error of the indirect distance function.

[0028] Further preferably, a dynamic adjustment mechanism for the weight coefficient between the next moment and the current moment is set according to the following formula;

[0029]

[0030] wherein, is the learning rate, is the weight factor of the direct error, is the weight factor of the indirect error; is the root mean square of the historical error of the sensor; is the working environment change coefficient; are respectively the influence factors of the historical error and the environment change on ;

[0031] Further preferably, the learning rate varies dynamically with the historical stability of the sensor and is dynamically adjusted according to the historical error feedback;

[0032]

[0033] where is the initial learning rate;

[0034] is the root mean square of the historical error of the sensor; is the parameter controlling the attenuation of the learning rate. When is larger, it indicates that the stability of the sensor is worse, and is reduced to avoid frequent adjustment; when is smaller, it indicates that the stability of the sensor is better, and is increased to improve the adjustment speed.

[0035] The elevator digital twin virtual and physical sensor data calibration method disclosed in the present application has at least the following advantages compared with the prior art:

[0036] 1. In the present application, the measured data of the physical sensor is combined with the virtual sensor measurement data of the digital twin. The sensor deviation is calculated and calibrated in real time, which can effectively solve the problem of single sensor error. By dynamically selecting the distance function and adaptively adjusting the calibration tolerance, this method realizes the comprehensive coordination of different sensor types and ensures the accuracy of the sensor data during the elevator operation.

[0037] 2. The present invention uses the direct distance function and the indirect distance function for fusion calibration. By monitoring the states of the key components of the elevator, the calibration can be automatically adjusted according to the deviation of the sensor data. This method not only improves the fault tolerance ability of the elevator, but also can discover potential problems in advance through historical data analysis and trend prediction, reduce the downtime and sudden failures, and improve the reliability of the long-term monitoring of the elevator. Brief Description of the Drawings

[0038] Figure 1 This is a schematic structural diagram of a method for calibrating data of virtual and physical sensors in a digital twin of an elevator according to the present invention.

[0039] Figure 2 This is a schematic flow diagram of calibrating virtual data of a digital twin by an indirect distance function according to the present invention. Detailed Embodiments

[0040] The present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0041] As Figure 1 shown, a method for calibrating data of virtual and physical sensors in a digital twin of an elevator provided by an embodiment of the present invention on the one hand includes the following steps:

[0042] First, perform a reference environment check; define the reference environment: during the normal operation of the elevator, select a standard operating condition as the calibration reference environment. The reference environment should meet the following conditions:

[0043] The elevator is in a stable operating state without abnormal faults.

[0044] Both the physical sensor and the virtual sensor are within the working range and the data is stable.

[0045] Environmental factors (such as temperature, humidity, vibration, etc.) should be close to normal operating conditions to avoid excessive interference with the sensor data.

[0046] Data acquisition: It is necessary to arrange various types of sensors (such as temperature sensors, current sensors, acceleration sensors, etc.) in the elevator system to collect sensor data in the reference environment. Based on the above data, a calibration reference model is constructed to provide a standard reference for subsequent error calibration.

[0047] S1. Obtain real-time data of various physical sensors in the elevator system in the reference environment and real-time data of virtual sensors in the elevator digital twin model at the same moment t;

[0048] Physical sensors such as temperature sensors, acceleration sensors, current sensors, etc., and virtual sensors are based on the data calculated by the elevator simulation model.

[0049] S2. Preprocess the data. Preprocess the data, compare the collected data of the same type of physical sensor and virtual sensor according to the distance function, and obtain the error D fused (t) between the physical sensor and the virtual sensor at moment t;

[0050] Among them, the preprocessing includes: removing noise, filling in missing values, and performing time synchronization; comparing the data of the acquired physical sensors and virtual sensors at the same timestamp; it also includes error type identification: according to the sensor type, possible errors include drift error, systematic error, error caused by environmental changes, etc. When calculating the error, it is necessary to combine the working characteristics of the sensors to identify the error type.

[0051] Compare the data of the same type of physical sensors and virtual sensors collected at the same moment t according to the distance function to obtain the error between the physical sensor and the virtual sensor at moment t D fused (t);

[0052] Definition of the direct distance function: used to calculate the error difference between the physical sensor and the virtual sensor under the same operating conditions. Applicable to sensors in the same or similar working environments.

[0053] Definition of the indirect distance function: used to handle the scenario of data fusion between multiple physical sensors and virtual sensors, and calculate the overall error by analyzing the error transfer relationship between multiple sensors.

[0054] The direct distance function focuses on solving the errors of physical sensors. By calculating the error between the physical sensor data and the predicted value of the virtual sensor, the direct distance function helps to identify and correct the deviation of the physical sensor. Because physical sensors are easily affected by external factors (such as temperature changes, electromagnetic interference, etc.) and exhibit drift or errors. The specific process is as Figure 1 shown.

[0055] The direct distance function is expressed by the following formula:

[0056]

[0057] Among them, is the error between the physical sensor and the virtual sensor calculated at moment ; is the physical sensor data, is the virtual sensor data.

[0058] Collect the data of the physical sensor and the virtual sensor at the same time point; calculate the difference between the two to obtain the error between the physical sensor and the virtual sensor calculated at moment ; and compare the error with whether it exceeds the set error threshold . If the error exceeds the threshold, a calibration operation is performed to adjust the physical sensor data through linear regression and offset correction functions.

[0059] S3. Comparison error D fused (t) Whether it exceeds the set error threshold ; If the error exceeds the threshold, the physical sensor data is corrected through linear regression calibration and offset correction function, and the virtual sensor data at the next moment in the digital twin is updated.

[0060] Linear regression calibration: Correct the error of the physical sensor by establishing a linear relationship between the physical sensor data and the virtual sensor data: ;

[0061] in, and is the regression coefficient estimated by the least squares method. This calibration function adjusts the physical sensor data Bringing it closer to virtual sensor data .

[0062] Offset correction: If the error is mainly offset error (such as long-term drift of the sensor), the offset correction function can be used to correct the offset; ;

[0063] Select the linear regression calibration method or the offset correction method based on the gradient of the difference between the previous and next acquisition moments. Because the gradient change reflects the state evolution of the sensor between consecutive moments, if the sensor error presents a stable offset, the gradient change is small, and offset correction can be used first. If the gradient changes significantly and shows a linear trend, it may be caused by gain error or external disturbance, and linear regression is more suitable at this time. If the gradient changes dramatically and irregularly, it indicates that there may be external noise, interference or hardware failure.

[0064] Real-time feedback: Confirm that the physical sensor data calibration is complete, and the corrected data will be fed back to the elevator system in real time.

[0065] When updating the virtual sensor data at the next moment in the digital twin, the virtual sensor model is adjusted using the weighted average method based on the error calculated by the indirect distance function: ;

[0066] in: is the weight of the physical sensor, is the weight of the virtual sensor, and the error between the physical sensor and the virtual sensor is minimized by optimizing the weighting coefficient. It is the weighted average of virtual sensors and physical sensors.

[0067] The indirect distance function focuses on globally optimizing all sensors in the system by analyzing data from multiple sensors (a combination of physical sensors and virtual sensors). Even if the error of each sensor is relatively small, the combined error among multiple sensors may affect the accuracy of the entire elevator system. Therefore, the indirect distance function can find the error transfer relationship among multiple sensors and further optimize the entire system. The specific process is as Figure 2 shown. Based on the indirect distance function, a multi-sensor error model is established using the obtained data from multiple physical sensors, virtual sensors, and the predicted values of virtual sensors. The obtained errors are weighted and averaged for optimization to minimize the errors. Finally, the elevator digital twin system is updated in real-time to update the virtual data.

[0068] The aforementioned indirect distance function is expressed by the following formula:

[0069] ;

[0070] where: is the predicted value of the physical sensor based on the virtual sensor ; n is the number of physical sensors, and m is the number of virtual sensors. is the error calculated between the predicted value and the physical sensor at time ; i is the index of the actual physical sensor data set, and j represents the index of the corresponding virtual sensor data set.

[0071] Among them, the predicted value is obtained by predicting the data collected by various physical sensors using a neural network established based on the correlation between sensors. The obtained

[0072]

[0073] where NN() represents the neural network; after inputting the input value V 1 (t), etc. into the neural network, f(V j (t)) is the obtained predicted value.

[0074] In the elevator system, the single-sensor error and the multi-sensor error may have different influence weights. To integrate these two error models, in S4, the error compared with the threshold is the final fused error obtained by weighted fusion of the direct distance function and the indirect distance function D fused ;

[0075] ;

[0076] where: is the error function after fusion; is the weight coefficient; is the direct distance function, is the error of the indirect distance function.

[0077] Further preferably, a dynamic adjustment mechanism for the weight coefficient between the next moment and the current moment is set according to the following formula

[0078]

[0079] wherein, is the learning rate, is the weight factor of the direct error, is the weight factor of the indirect error; is the root mean square of the historical error of the sensor; is the working environment change coefficient; are respectively the influence factors of the historical error and the environment change on ;

[0080] Further preferably, the learning rate varies dynamically with the historical stability of the sensor and is dynamically adjusted according to the historical error feedback,

[0081]

[0082] wherein is the initial learning rate; is the root mean square of the historical error of the sensor; is the parameter for controlling the attenuation of the learning rate. When is larger, it indicates that the stability of the sensor is worse, and is reduced to avoid frequent adjustment; when is smaller, it indicates that the stability of the sensor is better, and is increased to improve the adjustment speed.

[0083] Obviously, the above embodiments are only examples for clear illustration and are not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation manners here. And the obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims

1. A method for calibrating virtual and physical sensor data of an elevator digital twin, characterized in that: include: S1. Obtain real-time data of multiple types of physical sensors in the elevator system under the reference environment at the same time t and real-time data of virtual sensors in the elevator digital twin model; S2. Preprocess the data, compare the collected physical sensor and virtual sensor data of the same category according to the distance function, and obtain the error between the physical sensor and the virtual sensor at time t. D fused (t); The distance function includes a direct distance function and an indirect distance function. The direct distance function is used to calculate the error between the physical sensor data and the virtual sensor prediction value, and the indirect distance function is used to calculate and verify the matching relationship between each sensor and calibrate the transfer error; The direct distance function is expressed by the following formula: ； in, It is at the moment The calculated error between the physical sensor and the virtual sensor; is the physical sensor data, is virtual sensor data; The indirect distance function is expressed by the following formula: ； in: Based on virtual sensor data For physical sensor data The predicted value; n is the number of physical sensors, m is the number of virtual sensors; It is at the moment Computing predicted values ​​with physical sensors The error between them; i is the actual physical sensor dataset index, j is the corresponding virtual sensor dataset index; S3, comparison error D fused (t) Whether it exceeds the set error threshold ; If the error exceeds the threshold, the physical sensor data is corrected through linear regression calibration and offset correction function, and the virtual sensor data at the next moment in the digital twin is updated.

2. The elevator digital twin virtual and physical sensor data calibration method according to claim 1 is characterized in that: It also includes building a neural network to predict the data collected by various physical sensors based on the correlation between sensors. ； Among them, NN represents neural network, f(V j (t)) Based on virtual sensor data For physical sensor data The predicted values ​​of V1(t), V2(t), V m (t) is the neural network input value, when When j takes different index numbers of 1, 2, and m Corresponding virtual sensor data.

3. The elevator digital twin virtual and physical sensor data calibration method according to claim 1 is characterized in that: In S3, when updating the virtual sensor data at the next moment in the digital twin, the virtual sensor model is adjusted using the weighted average method based on the error calculated by the indirect distance function: ; in: is the weight of the physical sensor, is the weight of the virtual sensor, and the error between the physical sensor and the virtual sensor is minimized by optimizing the weighting coefficient; It is the weighted average of virtual sensors and physical sensors.

4. The elevator digital twin virtual and physical sensor data calibration method according to claim 1 is characterized in that: In S2, when the comparison is performed according to the distance function, the distance function is a result of weighted fusion of the direct distance function and the indirect distance function, which is expressed as follows: ； in: is the error function after fusion; is the weight coefficient; is the direct distance function, is the error of the indirect distance function.

5. The elevator digital twin virtual and physical sensor data calibration method according to claim 4 is characterized in that: The dynamic adjustment mechanism of the weight coefficient between the next moment and the current moment is set according to the following formula; in, is the learning rate, is the weight factor of the direct error, is the weight factor of indirect error; is the historical RMS error of the sensor; is the coefficient of variation of the working environment; are historical errors and environmental changes. The impact factor.

6. The elevator digital twin virtual and physical sensor data calibration method according to claim 5 is characterized in that: The learning rate Dynamically adjust based on historical error feedback as the sensor's historical stability changes; in is the initial learning rate; is the historical RMS error of the sensor; is the parameter that controls the learning rate decay. The larger the value, the worse the sensor stability. To avoid frequent adjustments; The smaller the value, the better the stability of the sensor. Improve adjustment speed.

Images