DATA CALIBRATION METHOD FOR ELEVATOR DIGITAL TWIN VIRTUAL AND PHYSICAL SENSORS

Abstract

U I T T R E K S E L The invention discloses a data calibration method for elevator digital twin virtual and physical sensors. It includes obtaining the real—time data of multiple types of physical sensors in the elevator system under the benchmark environment at the same time t and real—time data of virtual sensors in the elevator digital twin model; preprocessing the data, and comparing collected data of the same type of physical sensor and virtual sensor according to a distance function to obtain an 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, correcting the physical sensor data by linear regression calibration and offset correction function, and updating the virtual sensor data at the next moment in the digital twin. This application combines the measured data of the physical sensor with the calculated. data of the virtual sensor of the digital twin to calculate and calibrate the sensor deviation in real—time, which can effectively solve the problem of singlesensor error. (+ Fig. l)

Description

P2128 / NL DATA CALIBRATION METHOD FOR ELEVATOR DIGITAL TWIN VIRTUAL AND PHYSICAL SENSORS TECHNICAL FIELD The invention relates to the field of elevator digital twin technology, in particular to a data calibration method for eleva tor digital twin virtual and physical sensors. BACKGROUND With the rapid development of China's economic construction and the rapid construction of tall buildings, the elevator has be come an indispensable facility in people's daily production and life. It is a kind of special equipment directly related to the safety of people's lives and properties. At present, with the widespread popularity of the Internet of Things, the elevator is also gradually applied to the Internet of Things system, but due to a variety of uncertainties of the system, and the physical sen- sors of the elevator in the long-term operation of the instability also exists. Because the hardware unit of the physical sensor, the communication unit, and the environment around the sensor may cause the sensor to be abnormal and produce noise, offset, drift, and other phenomena. For example, there is a sensor with intermit tent drift error in the elevator sensing system (traction machine temperature, vibration sensor), due to the compact deployment space of the sensor and the surrounding control cabinet thermal radiation interference, the temperature sensor will produce tem- perature and vibration deviation due to electromagnetic interfer- ence, thermal radiation, hardware anomalies, and other factors, so when faced with a limited and unreliable sensing environment, it will harm the digital twin system in the elevator, the datadriven applications in elevator digital twin systems are negatively im pacted when faced with a limited and unreliable sensing environ ment. The digital twin technology integrates multiphysical, mul tidisciplinary, and multiscale attributes, with the characteris tics of hyperrealistic, realtime synchronization, allelement, and uniqueness, which can realize the interactive integration of the physical world and the information world, so as to reflect the entire lifecycle process of the corresponding physical equipment. At present, the retrieved digital twin virtual and physical consistency method is the patent of Complex System Level Digital Twin Operation virtual and physical consistency Determination and Interaction Method (Application No. 202211410916.5), which dis closes a complex system level digital twin operation virtual and physical consistency determination and interaction method, de scribes the subspace in the virtual and physical space of the com plex system, and proposes a method for determining the virtual and physical consistency based on the dynamic operation, and proposes a specific operation and an interaction method between the virtual and physical subspaces; The method establishes the operation and interaction process between the virtual space and the physical space according to the association relationship of the digital twin of the complex sys tem, but there are still some shortcomings: l. The technique has not yet considered how to ensure the ac- curacy and timeliness of the data transmitted between the physical and virtual sensors, which can not further ensure the effective- ness of the digital twin virtual and physical models feeding each other. 2. The traditional methods usually focus on the error cali bration of a single sensor and the optimization of sensor calibra tion in the local or global environment of the digital twin is not yet considered. Therefore, there is a requirement for a data calibration method to ensure the reliability of physical and virtual sensors to ensure the accuracy and safety of the elevator digital twin un der the full life cycle operation, therefore, the invention pro poses a data calibration method for elevator digital twin virtual and physical sensors. SUMMARY Therefore, the purpose of this invention is to provide a data calibration method and a system for elevator digital twin virtual and physical sensors to solve the error problem between sensor da ta and ensure the consistency of virtual sensor and physical sen sor data. In order to achieve the above purpose, the invention provides a data calibration method and a system for elevator digital twin virtual and physical sensors, including the following steps: 81, obtaining realtime data of multiple types of physical sensors in an elevator system under a benchmark environment at the same time t and realtime data of virtual sensors in the elevator digital twin model; SZ, preprocessing the data, and comparing collected data of the same type of physical sensor and virtual sensor according to a distance function to obtain an error DmaMt) between the physical sensor and the virtual sensor at time t; S3, comparing whether the error DMt) exceeds a set error threshold 3; if the error exceeds the threshold, correcting the physical sensor data by linear regression calibration and offset correction function, and updating the virtual sensor data at the next moment in the digital twin. Furthermore, 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 phys ical 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 trans fer error. Furthermore, the direct distance function is expressed by the following formula: D (t) = IP(t) V<z)| _ where 1%Ma0) is the error between the physical sensor and the virtual sensor calculated at time t;PU) is the physical sen sor data and VKO is the virtual sensor data. Furthermore, the indirect distance function is expressed by the following formula: D (r) = Z Z lea) f(V, (ml i=1 j=1 . where ÍÏVÁÙ) is a predicted value of the physical sensor B(Û based on the virtual sensor DE); n is a count of the physical sensors, and m is a count of the virtual sensors. Furthermore, it also includes the construction of a neural network to predict the data collected by various physical sensors according to the correlation between sensors, the following formu la is obtained, f (V,» (t)) = NN (V1 (t), V2 (î); - - V. (t)) where NN denotes the neural network, f(Vj(t)) is the predict ed value; Vj(t), V2(t),V.(t) are input values. Furthermore, in 83, when updating the virtual sensor data at the next moment in the digital twin, the weighted average method is used to adjust the virtual sensor model according to the error calculated by the indirect distance function: Pcalibrated (t) = Zai Pi(t) + Zp; f(I / j(t)) i=1 j:] . where ai is a weight of the physical sensor, [Ë 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. Furthermore, in S2, when comparing according to the distance function, the distance function is the result of the weighted fu sion of the direct distance function and the indirect distance function, which is expressed according to the following formula; Djùsed (t) = Vl 'Ddirecz(t)+(1_ W) ' Dindirect(t) ; where [%(Û _is the error function after fusion; Vl is the weight coefficient; [%WÜÚ) is the direct distance function, [%WWÜU) is the error of the indirect distance function. Furthermore, the dynamic adjustment mechanism of the weight coefficient at the next moment and the current moment is set ac cording to the following formula; V / (Ï + 1) = WU) + n(WdDdírect _ WiDindirecz) + A ' Eh + ' Ce where Ü is the learning rate, MQ is the weight factor of the direct error, vg is the weight factor of the indirect error; la is the root mean square of the historical error of the sensor; C; is the coefficient of variation of working environment; ! 8 are the influence factors of the historical error and the environmental change respectively. Furthermore, the learning rate n changes dynamically with the historical stability of the sensor, and is dynamically adjusted according to the historical error feedback. 77 = 77 "'1 ° ;.E, where no is the initial learning rate; Ed is a root mean square of the historical error of the sen sor; 4, is a parameter that controls the attenuation of the learn ing rate, when ET is larger, the stability of the sensor is worse, and n is reduced to avoid frequent adjustment, when the value of [% is smaller, the stability of the sensor is better, and the in crease of n improves the adjustment speed. The data calibration method for elevator digital twin virtual and physical sensors in this application has at least the follow ing advantages compared with the existing technology: 1. In this application, the measured data of physical sensors are combined with the virtual sensor measurement data of digital twins. Realtime calculation and calibration of sensor deviation can effectively solve the problem of single sensor error. By dy namically selecting the distance function and adaptively adjusting the calibration tolerance, the method realizes the comprehensive coordination of different sensor types and ensures the accuracy of sensor data in elevator operation. 2. The fusion calibration is carried out by using the direct distance function and the indirect distance function, by monitor ing the state of the key components of the elevator, the calibra tion can be automatically adjusted according to the deviation of the sensor data. This method not only improves the fault tolerance of elevators, but also can find potential problems in advance through historical data analysis and trend prediction, reduce downtime and sudden failures, and improve the reliability of long term elevator monitoring. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a structural diagram of the data calibration method for the elevator digital twin virtual and physical sensors in the invention. FIG. 2 is a flow diagram of the indirect distance function correcting digital twin virtual data in the invention. DETAILED DESCRIPTION OF THE EMBODIMENTS The following is a further detailed description of the inven tion through drawings and specific embodiments. As shown in FIG. 1, on the one hand, the embodiment of the invention provides a data calibration method for elevator digital twin virtual and physical sensors, including the following steps: First, the benchmark environment is verified; the reference environment is defined: During the normal operation of the eleva tor, a standard operating condition is selected as the calibration reference environment. The benchmark environment should meet the following conditions: The elevator is in stable operation and there is no abnormal fault. Both the physical sensor and the virtual sensor are within the working range and the data is stable. Environmental factors (such as temperature, humidity, vibra tion, etc.) should be close to normal working conditions to avoid excessive interference with sensor data. Data acquisition: It is necessary to arrange multiple types of sensors (such as temperature sensors, current sensors, acceler ation sensors, etc.) in the elevator system to collect sensor data in the reference environment. Based on the above data, a calibra tion benchmark model is constructed to provide a standard refer ence for subsequent error calibration. Sl, the realtime data of multiple types of physical sensors in the elevator system under the benchmark environment at the same time t and the realtime data of the virtual sensors in the eleva tor digital twin model are obtained; the physical sensors such as temperature sensors, accelera tion sensors, current sensors, etc., and virtual sensors are based on the data calculated by the elevator simulation model. 82, the data is preprocessed, and the data is preprocessed, the collected data of the same type of physical sensor and virtual sensor are compared according to the distance function to obtain the error Dat) between the physical sensor and the virtual sen sor at time t; the preprocessing includes removing noise, filling in missing values, and time synchronization; the data of the acquired physi cal sensor and the virtual sensor are compared under the same timestamp; it also includes error type identification: Possible errors include drift error, system error, error caused by environ mental changes, etc. according to the type of sensor. When calcu lating the error, it is necessary to combine the working charac teristics of the sensor to identify the error type. The data of the same type of physical sensor and virtual sen sor collected at the same time t are compared according to the distance function, and the error DMt) between the physical sen sor and the virtual sensor at time t is obtained; The definition of the direct distance function: It is used to calculate the error difference between the physical sensor and the virtual sensor under the same operating conditions. It is suitable for sensors in the same or similar working environment. The definition of the indirect distance function: It is used to deal with the scene of data fusion between multiple types of physical sensors and virtual sensors, and calculate the overall error by analyzing the error transfer relationship between multi ple sensors. The direct distance function focuses on solving the error of the physical sensor. By calculating the error between the physical sensor data and the predicted value of the virtual sensor, the di rect distance function helps to identify and correct the deviation of the physical sensor. Because the physical sensor is easily af fected by external factors (such as temperature change, electro magnetic interference, etc.), drift or error occurs. The specific process is shown in FIG. 1. The direct distance function is expressed by the following formula: D... (t) = |P(t) V(t)| where [%WÜÜ) is the error between the physical sensor and the virtual sensor calculated at time t; PU) is the physical sen sor data, and VU) is the virtual sensor data. The data of the physical sensor and virtual sensor at the same time point are collected; the difference between the two sen sors is calculated, and the error LQWMU) between the physical sen sor and the virtual sensor calculated at time t is obtained, and whether the error [%WÜU) exceeds the set error threshold 5 is compared. If the error exceeds the threshold, the calibration op eration is performed, and the physical sensor data is adjusted by linear regression and offset correction function. 83, whether the comparison error Dat) exceeds the set er ror threshold 3 is compared; if the error exceeds the threshold, the physical sensor data is corrected by linear regression cali bration and offset correction function, and the virtual sensor da ta at the next moment in the digital twin is updated. Linear regression calibration: The error of the physical sen sor is corrected by establishing a linear relationship between the physical sensor data and the virtual sensor data: Ealíhmted (t) = PO) + OC ' Ddríect (î)+ ; where d and 0 are the regression coefficients estimated by the least square method. The calibration function makes it closer to the virtual sensor data VU) by adjusting the physical sensor data PU). Offset correction: If the error is mainly offset error (such as longterm drift of the sensor), the offset correction function can be used to correct the offset; FQMWW(ÛIIPU)lìmm(Û; the linear regression calibration method or the offset cor rection method is selected according to the gradient change of the difference between the acquisition time before and after; because the gradient change reflects the state evolution of the sensor be tween continuous moments, if the sensor error shows a stable off set, the gradient change is small, and the offset correction can be used preferentially; if the gradient changes significantly and linearly, it may be caused by gain error or external disturbance, then linear regression is more suitable; if the gradient changes drastically and irregularly, there may be external noise, inter ference, or hardware failure. Realtime feedback: whether the physical sensor data correc tion is completed is confirmed, and the corrected data will be fed back to the elevator system in realtime. When updating the virtual sensor data at the next moment in the digital twin, the weighted average method is used to adjust the virtual sensor model according to the error calculated by the " m Ealibrated (Ï) = Z. 'PzÜ) + ZIBÍ "f(Vj @» indirect distance function: l j ; where ai 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. The indirect distance function focuses on the global optimi zation of all sensors in the system by analyzing the data of mul tiple sensors (a combination of physical sensors and virtual sen sors). Even if the error of each sensor is relatively small, the comprehensive error between multiple sensors may affect the accu racy of the entire elevator system, therefore, the indirect dis tance function can find the error transfer relationship between multiple sensors and further optimize the entire system. The spe cific process is shown in FIG. 2. The obtained multiple groups of physical sensor data and virtual sensor data, as well as the pre dicted value of the virtual sensor, are used to establish a multi sensor error model based on the indirect distance function. The obtained error is weighted and averaged to minimize the error, fi nally, the elevator digital twin system is updated in realtime, and the virtual data is updated. The indirect distance function. It is expressed by the fol lowing formula: n m D (r) = Z Z|B <r>f(V, (>| ll ; where f(VÄfD is the predicted value of the physical sensor B(Û based on the virtual sensor PZG); n is the count of the phys ical sensors, and m is the count of the virtual sensors. The predicted value, according to the correlation between the sensors, the neural network is established to predict the data collected by various physical sensors. f(V,(t)) = NNU / 10), Vzw),- - - V. (t)) where NN() denotes the neural network; after inputting the input value V1(t) into the neural network, f(V5(t)) is the predict ed value. In the elevator system, singlesensor error and multisensor error may have different influence weights. In order to integrate these two error models, in S4, the error compared with the thresh old is the final fusion error Dmafafter the weighted fusion of the direct distance function and the indirect distance function. D... (r) = w D (r) + (1 w) - D (r) ,- where [%MÚU) is the error function after fusion; V is the weight coefficient; I...O) is the direct distance function, [%MËÜU) is the error of the indirect distance function. Furthermore, according to the following formula, the dynamic adjustment mechanism of the weight coefficient of the next moment and the current moment is set W(l + 1) = ((I) + n(WdDdirect _ wiDinafirect) + lu . Eh + 8 . Ce where Ü is the learning rate, MQ is the weight factor of di w. . . . . E . rect error, 1 is the weight factor of the indirect error; h is the root mean square of the historical error of the sensor; C; is the coefficient of variation of the working environment; A are the influence factors of the historical error and environmental change on V, respectively. Furthermore, the learning rate n changes dynamically with the historical stability of the sensor and is dynamically adjusted ac cording to the historical error feedback. 77 77 + 1 ° 1+g~Eh where 0 is the initial learning rate; [% is the root mean square of the historical error of the sensor; 4, is the parameter that controls the attenuation of the learning rate, when Eh is larger, the stability of the sensor is worse, and n is reduced to avoid frequent adjustment, when the value of 5% is smaller, the stability of the sensor is better, and the increase of n improves the adjustment speed. Obviously, the above embodiment is only an example made to clearly explained, not to limit the embodiments. For the general technical personnel in their field, based on the above describe tion, other different forms of change or amendment can also be made, there is no need and can not be exhaustive of all the embod iments. The obvious changes or amendments thus extended are still within the scope of protection of the invention. CONCLUSIONS l. Calibration Method for Digital Dual Virtual and Physical sensor data from the elevator, characterized by having the following steps include: Sl: Acquires real-time data from multiple types of physical sensors in the elevator system and real-time data from virtual sensors di digital twin model of the elevator under the reference environment at the same moment t; S2: Preprocess the data and compare the collected data of physical and virtual sensors of the same type according to a position function to avoid an error Dmat) between the physical and virtual sensors available at the time; S3: Check whether the error Dmadt) is within a set error threshold exceeds; If the error exceeds the threshold, the phy sical sensor data corrected by linear regression calibration tie and offset correction functions to correct the virtual sensor data to update the next moment in the digital twin model. 2. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 1, characterized in that in S3, the distance function is a direct distance function and an indi includes the direct distance function, the direct distance function is called used to detect the error between the physical sensor data and the virtual to calculate sensor prediction value, the indirect distance function tie is used to establish the corresponding relationship between each sensor to calculate and verify, and to calibrate the transfer error run. 3. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 2, characterized in that the direct distance function is represented by the following formula: D (r) = |P(t) Vol ; where [@@ÜÜ) is the error between the physical sensor and the virtual one sensor calculated at time t; PU) is physical sensor data, and VTÛ is virtual sensor data. 4. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 2, characterized in that the indirect distance function is represented by the full following formula: D (r) = Z Z IP,-(t) f(V, <r>>| i:] j:] . V. t . where f(f()) is the predicted value of physical sensor B(Û ge . V. (t) . . based on Virtual sensor ] ; n is the number of physical sensors, and m is the number of virtual sensors. 5. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 4, characterized in that it further includes setting up a neural network based on the correlation between sensors to predict the data that are collected by various physical sensors, and obtain f(V,-(t)) = Nwa), V20),- - - V (t)) _ where NN stands for neural network, and f(Vj(t)) is the obtained pin value; Vï(t), V2(t), Vh(t) are input values. 6. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 2, characterized in that in S3, when updating the virtual sensor data at the next time in the digital twin model, the virtual sensor model The share is adjusted using the weighted average method based on the error calculated by the indirect distance function: Ralibrated (t) = Z ai . Pi(t) + Z ßi ' f(Vj (t)) i:] j:] . . . . a. . . ß < . . where is the weight of the physical sensor, J 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. 7. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 4, characterized in that in SZ, when compared according to the distance function, the distance function is the result of weighted fusion of the direct derivation position function and the indirect distance function, represented by the following formula: Dfused (t) = W . Ddirect(t)+(1_ W) . Dindirect (t) . ... D(z) . . w . . .. where « is the off-function after merger; is a weight coefficient ficient; l%WÚÜ) is a direct distance function and [%WU) is the error of an indirect distance function. 8. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 7, characterized in that a dynamic adjustment mechanism for the weight coefficients of the next time and the current time is set according to the following formula: ((tFl) : l / I(t)+î7(WdDdirect WiDindirect)+lu.Eh + . C8 where Ü is the learning rate, ® is the weight factor of the di w . . . . E . direct error, l is the weight factor of indirect error; h is the middle square of the sensor's historical error; C; is the coefficient of change in the work environment; Human representation respectively closes the impact factors of historical errors and environmental changes on V} 9. Calibration method for digital dual virtual and physical sensor data of the elevator according to claim 8, characterized in that the learning rate n changes dynamically with the historical stability sensor and is dynamically adjusted based on its toric error feedback: 77 _ _ 77 o "1 1+Ç-Eh where Üo is the initial learning rate; [& is the middle four side of the sensor's historical error; gf is a parameter which controls the decay of learning rate, and the larger h is, the worse the stability of the sensor is, and currently, currently reduce n to avoid frequent adjustments; the Smaller h is, the better the stability of the sensor is, and this moment, increase n to improve the adaptation rate. S1 Obtaining the real-time data of multiple types of physical sensors in the elevator system under the benchmark environment at the same time t and real-time data of virtual sensors in the elevator digital twin model S2 processing the data, and comparing collected data of the same type of physical ensor and virtual sensor according to the distance function to obtain the error D (t) between the physical sensor and the virtual sensor at time t fused S3 omparing whether the error D (t) exceeds the set error threshold; if the error fused xceeds the threshold, correcting the physical sensor data by linear regression bration and offset correction function, and updating the virtual sensor data at the next moment in the digital twin FIG.1 Calibrating reference environment Predicted virtual Virtual sensor 1 Physical sensor 1 data 1 Predicted virtual Physical sensor 2 Virtual sensor 2 data 2 ... ... ... Predicted virtual data Virtual sensor m Physical sensor n m Establishing a multi-sensor error model based on the indirect distance function Error correction Minimizing the error by optimizing the weighting coefficient Real-time feedback to update the elevator digital twin system FIG.2< / r> < / r>

Images