A total station self-learning measurement adjustment method and system based on rolling prediction

Abstract

The application is a total station self-learning measurement adjustment method and system based on rolling prediction. It includes a gateway control module, a data acquisition and processing module, and a platform management module. The gateway control module is connected with the data acquisition and processing module and the platform management module respectively, responsible for monitoring equipment control and data transfer storage; the data acquisition and processing module is responsible for collecting monitoring data, and taking the coordinates predicted based on the Transformer time series prediction model as the initial coordinates of indirect adjustment to reduce the iteration times of adjustment to obtain monitoring coordinates; the platform management module realizes remote management of equipment and visualization of data. The application uses predicted coordinates closer to true values as initial approximate coordinates of adjustment, which can accelerate the convergence of adjustment iteration. The monitoring point coordinates obtained by adjustment are used to update the rolling prediction model of the next observation period and to update the learning file; a self-learning closed loop with accurate model prediction, fast total station aiming, and stable adjustment results is formed, and the system performance is continuously optimized.

Description

Technical Field

[0001] This invention relates to the field of engineering deformation monitoring technology, specifically to a total station self-learning measurement adjustment method and system based on rolling prediction. Background Technology

[0002] Engineering deformation monitoring is a core technological means to ensure the construction safety and long-term stable operation of major engineering projects. Its core objective is to identify the deformation trend of engineering structures and provide early warnings of safety risks such as collapse and slippage by accurately acquiring changes in the spatial position of monitoring points. Total stations, as high-precision measuring instruments integrating angle and distance measurement functions, have become the mainstream observation equipment in the field of engineering deformation monitoring due to their advantages such as convenient operation, high measurement accuracy, and adaptability to complex engineering environments. Especially in automated monitoring scenarios, by pre-setting learning files (including aiming parameters for monitoring points), periodic, unattended continuous data acquisition can be achieved, significantly improving monitoring efficiency.

[0003] In the process of solving total station deformation monitoring data, three-dimensional indirect adjustment is the core method for processing observation data and obtaining accurate three-dimensional coordinates of monitoring points. This method uses the three-dimensional coordinates of the monitoring points as unknown parameters. Based on the geometric function relationship between the observed horizontal angles, vertical angles, and slope distances of the total station and the unknown parameters, an error equation is established through Taylor series linearization. Then, the corrections for the unknown parameters are solved according to the least squares criterion, ultimately obtaining the most probable coordinate values ​​for each monitoring point. To ensure calculation accuracy, when the initial approximate coordinates deviate significantly from the true coordinates, iterative adjustment is needed to correct the linearization error until the correction is less than the convergence threshold, completing the calculation. Simultaneously, the continuity of automated total station monitoring depends on the aiming parameters in the learning file, and their accuracy directly determines the aiming success rate and measurement accuracy in subsequent observations.

[0004] However, the existing total station deformation monitoring technology still has the following key technical defects in adjustment calculation and automated aiming coordination: (1) The prior information of deformation monitoring is not utilized, and the adjustment lacks specificity. In engineering deformation monitoring scenarios, the deformation direction of monitoring points generally has a clear regularity due to the influence of engineering structural characteristics, stress mechanism and geological environment. This deformation direction can be obtained in advance through engineering geological analysis, statistics of previous monitoring data, etc., and the predicted coordinates of the monitoring points can be obtained accordingly. However, the existing three-dimensional indirect adjustment method does not utilize such information and mostly uses the previous period coordinates or geometric approximations to determine the initial coordinate approximation. Such initial values ​​deviate greatly from the current period's true coordinates, resulting in the corrections being scattered in the X, Y and Z directions during adjustment iteration, which reduces the adjustment calculation efficiency and numerical stability. (2) The learning files for automated monitoring are static, and the reliability of total station aiming is insufficient. The learning files for existing total station automated monitoring are mostly generated by initial calibration, which only store the initial aiming parameters of the monitoring points and do not establish a linkage mechanism with the adjustment calculation results. As the engineering structure continues to deform, the angle and slant distance parameters in the initial learning file are no longer applicable to aiming at the deformed points, which will reduce the speed of the total station aiming the prism.

[0005] In summary, existing total station deformation monitoring technology has two major shortcomings: first, the adjustment calculation does not incorporate prior deformation, resulting in slow convergence and susceptibility to accuracy interference; second, the learning file for automated aiming lacks a dynamic update mechanism for deformation information, leading to insufficient aiming reliability. This lack of synergy between the two technologies means that existing technology cannot fully meet the triple requirements of engineering deformation monitoring: "efficient calculation, accurate measurement, and continuous automation." Therefore, an integrated technical solution that can combine predicted deformation values ​​with dynamically updated learning files is urgently needed. Summary of the Invention

[0006] To address the shortcomings of the aforementioned technologies, this invention provides a total station self-learning measurement adjustment method and system based on rolling prediction.

[0007] In a first aspect, the present invention provides a total station self-learning measurement adjustment method based on rolling prediction. This method is implemented through a total station self-learning measurement adjustment system based on rolling prediction, and includes the following steps:

[0008] S1: Set up the total station at the working base point, level the instrument, and fix the tripod.

[0009] S2: According to the direction observation method, take measurements by aiming at each deformation observation point and the known benchmark point in sequence, and save the learning file using an external data acquisition device: record the measurement data, including the horizontal angle, vertical angle and slope distance in each direction.

[0010] S3: Use communication equipment to modify the timed measurement task configuration file of the total station in the data acquisition instrument.

[0011] S4: According to the timed task configuration, perform timed measurements on the total station, and calculate the measurement error after each half-cycle. If the error exceeds the limit, the measurement cycle is repeated.

[0012] S5: The measurement results are processed using a three-dimensional indirect adjustment method. First, a time series model is introduced to predict the deformation of the monitoring point at the current moment. Based on this, the predicted coordinates of the monitoring point are calculated and directly set as the initial approximate coordinates required for the adjustment iteration. Then, based on the functional relationship between the total station observations and the coordinate parameters of the monitoring point, an error equation is established. On this basis, iterative calculation is performed using the least squares principle: each iteration solves for the correction of the approximate coordinate value and updates the approximate coordinate value accordingly. This process is repeated until the correction is less than the preset limit, at which point the iteration terminates. Finally, the observation correction value V and the coordinate approximate value correction value x are output synchronously. Then, the adjustment values ​​of the working base point coordinates, deformation monitoring point coordinates, observed horizontal angle, vertical angle, and slope distance, as well as the covariance matrix, are obtained.

[0013] S6: Unstable benchmarks can lead to distortion in the calculation of station coordinates. To ensure the reliability of the adjustment results, the average gap method is used to search for unstable points. Unstable points are eliminated and the adjustment is repeated. If the results still do not meet the requirements, the measurement is repeated and steps S4-S6 are repeated.

[0014] S7: The coordinates of each monitoring point calculated by adjustment are transmitted to the historical monitoring point database through the built-in program of the external data acquisition instrument, and the updated coordinate data are used to optimize and update the rolling prediction model parameters for the next cycle, forming a self-learning closed loop.

[0015] S8: Use the observations obtained from S5—horizontal angle, vertical angle, and slope distance—to update the learning data in the learning file; avoid inaccurate laser aiming due to deformation when the total station is used for measurement, and improve the success rate and aiming speed of measurement.

[0016] S9: Repeat steps S4-S8.

[0017] Further, in step S5, the deformation of the monitoring point at the current moment is predicted, specifically including: (1) obtaining the historical monitoring point coordinate database, calculating the cumulative change in each period, of which 80% is used as the training set, 10% as the validation set, and 10% as the test set; (2) obtaining the Transformer-based time series prediction model; training the model, using the validation set to fine-tune the model parameters, using the test set to evaluate the generalization of the model, and finally determining the model; (3) using the finally determined time series prediction model to infer the deformation of the monitoring point at the current moment based on the historical monitoring point coordinate database.

[0018] Furthermore, in step S5, the calculation method for the three-dimensional indirect adjustment is as follows:

[0019] The slope distance from the station point to n monitoring points can be obtained by freely setting up the total station. Horizontal angle vertical angle The coordinates of the survey station are: The coordinates of the monitoring point are If the orientation angle of the station is A, then the equation for the observed value is:

[0020] (1)

[0021] The component form of its linearized error equation is:

[0022] (2)

[0023] Where i represents the i-th monitoring point observed by the monitoring station. These are the approximate coordinates of the survey station. These are approximate coordinates of the monitoring station. It is the approximate spatial distance between the i-th measuring station and the monitoring point. Let be the approximate horizontal distance between the i-th measuring station and the monitoring point, and dx, dy, and dz be the corrections to the approximate coordinates of the measuring point, respectively. It is the correction value for the orientation angle of the measuring station. =206265, , , This is the constant term in the observation error equation; , , , , The formulas are as follows:

[0024] (3)

[0025] (4)

[0026] (5)

[0027] (6)

[0028] (7)

[0029] in, This is an approximate value of the orientation angle of the survey station;

[0030] Rewriting formula (2) in matrix form, we have:

[0031] (8)

[0032] in, , , , .

[0033] According to the least squares criterion, formula (8) must satisfy... The requirement can be obtained using the mathematical method of finding the free extrema of a function. Transpose Substituting formula (8) into the equation, we obtain the normal equation as follows: Seeking ; where P is the weight matrix of the observations.

[0034] Iterative adjustment process: judgment If k=0.1mm; if yes, output x, calculate the coordinates of each monitoring point based on the correction x, substitute x into formula (8) to obtain the observation correction V, and then obtain the adjusted values ​​of the observed horizontal angle, vertical angle and slope distance; if no, perform iterative solution until .

[0035] Secondly, the present invention also provides a total station self-learning measurement adjustment system based on rolling prediction, used to execute a total station self-learning measurement adjustment method based on rolling prediction as described in the first aspect, wherein the total station self-learning measurement adjustment system based on rolling prediction includes the following modules: a gateway control module, a data acquisition and processing module, and a platform management module; the gateway control module and the data acquisition and processing module are connected via an RS485 serial port, and the gateway control module establishes a network communication connection with the PC / Android platform management module via HTTP POST request / MQTT protocol.

[0036] Furthermore, a total station self-learning surveying adjustment system based on rolling prediction is provided, wherein the gateway control module undertakes the functions of equipment operation, data relay, and local data retention at the field end; the gateway control module includes three sub-functional modules: equipment control, data forwarding, and local storage; wherein:

[0037] The equipment control module is responsible for powering on / off the total station, tilt checking, triggering measurements, and scheduling timed measurement tasks.

[0038] The data forwarding module is responsible for receiving the raw measurement data returned by the total station, transmitting the data to be processed to the data acquisition and processing module, and forwarding the processed monitoring results to the platform management module.

[0039] The local storage module stores the original measurement data and the processed monitoring point coordinate data locally, providing support for data traceability.

[0040] Furthermore, a total station self-learning measurement adjustment system based on rolling prediction is provided, wherein the platform management module is responsible for realizing online management of the equipment and visualization analysis of total station monitoring point data.

[0041] Furthermore, a total station self-learning measurement adjustment system based on rolling prediction is provided, wherein the data acquisition and processing module is used to execute the total station self-learning measurement adjustment method based on rolling prediction described in the first aspect, and to complete the acquisition, processing and self-updating of measurement data and learning files.

[0042] Compared with the prior art, the present invention has the following advantages:

[0043] (1) Proposal of Rolling Prediction Method: This invention proposes a rolling prediction method based on a historical monitoring coordinate database. The prediction model is used to derive the coordinate values ​​of the current observation period, and the predicted coordinates are used as the initial approximation values ​​for three-dimensional indirect adjustment to perform adjustment calculations, and finally obtain the adjustment monitoring point coordinates at the current moment. At the same time, the coordinates are updated to the monitoring coordinate database for prediction and learning in the next measurement period. Compared with the traditional given initial value method (such as using the previous period coordinates or simple geometric estimation), the prediction method of this invention is based on the historical deformation trend for reasonable extrapolation, which is closer to the true coordinates at the current moment. It can reduce the number of iterations and avoid ill-conditioned problems of the normal equation matrix caused by poor initial values, thereby improving the stability of the solution process.

[0044] (2) Establishment of the total station self-learning mechanism: The present invention designs a self-learning mechanism between the total station learning file and the three-dimensional indirect adjustment results. By using the corrections obtained from the adjustment to correct the preset angle and distance values ​​in the learning file, the speed and accuracy of the total station aiming prism are significantly improved, and the measurement accuracy and efficiency are further optimized.

[0045] (3) Implementation of self-learning closed loop: This invention fully considers the diversity of changes in monitoring points and uses approximate values ​​that are closer to the real coordinates for indirect adjustment iteration, which significantly improves the calculation efficiency. The adjustment results are not only used to update the rolling prediction model for the next measurement cycle, but also fed back to the learning file to form a self-learning closed loop, thereby achieving continuous optimization of model prediction accuracy, total station aiming speed, and adjustment result stability, and improving the overall performance of the system. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating a total station self-learning measurement adjustment method based on rolling prediction proposed in this invention.

[0047] Figure 2 This is a flowchart of the approximate coordinate iterative process during the three-dimensional indirect adjustment calculation.

[0048] Figure 3 This is a schematic diagram of the structure of a total station self-learning measurement adjustment system based on rolling prediction proposed in this invention. Detailed Implementation

[0049] The following is in conjunction with the appendix Figure 1-3 The present invention provides a preferred embodiment and further illustrates the invention.

[0050] Example 1, as Figure 1 As shown, this invention provides a total station self-learning measurement adjustment method based on rolling prediction. This method is executed through a total station self-learning measurement adjustment system based on rolling prediction, and includes the following steps:

[0051] S1: Select a suitable location, set up the total station at the working base point, level the instrument, and fix the tripod; connect the total station and the data acquisition unit to form a network.

[0052] S2: According to the direction observation method, take measurements by aiming at each deformation observation point and the known benchmark point in sequence, and save the learning file using an external data acquisition device: record the measurement data, including the horizontal angle, vertical angle and slope distance in each direction.

[0053] S3: Use communication equipment to modify the timed measurement task configuration file of the total station in the data acquisition instrument.

[0054] S4: Upon receiving the timed measurement command from the gateway, measure the monitoring points sequentially according to the learning file, and transmit the measurement data to the gateway device in real time. After each half-cycle, calculate the line of sight error, vertical circle index difference, and cycle difference. If the limits are exceeded, the cycle is re-measured; otherwise, proceed to step S5.

[0055] S5: Perform three-dimensional indirect adjustment on the data obtained from the measurements in step S4:

[0056] (1) Obtain the historical monitoring point coordinate database and calculate the cumulative change in each period; 80% of the data is used as the training set, 10% as the validation set, and 10% as the test set.

[0057] (2) Construct a Transformer-based time series prediction model: Use deep learning frameworks (such as PyTorch, TensorFlow, etc.) to build a Transformer-based short-term time series prediction model. The model input is the cumulative change of historical monitoring points, and the output is the cumulative change of monitoring points in the next period.

[0058] (3) Input the training set into the prediction model for training, use the validation set to fine-tune the model parameters, and use the test set to evaluate the generalization of the model. By adjusting the parameters such as the model structure, optimizer, and loss function, the time series prediction model with the best evaluation index is finally obtained. The model is used to predict the change of the monitoring point at the current time, and the coordinates of the monitoring point are derived as the approximate initial coordinates in the indirect adjustment.

[0059] (4) such as Figure 2 As shown, first, the error equation is listed:

[0060] (1)

[0061] According to the least squares criterion, formula (1) must satisfy... The requirement can be obtained using the mathematical method of finding the free extrema of a function. Transpose Substituting formula (1) into the equation, we obtain the normal equation as follows:

[0062] (2)

[0063] in, , , , P is the weight matrix of the observations; These are the approximate coordinates of the survey station. These are approximate coordinates of the monitoring station. It is the approximate spatial distance between the i-th measuring station and the monitoring point. Let be the approximate horizontal distance between the i-th measuring station and the monitoring point, and dx, dy, and dz be the corrections to the approximate coordinates of the measuring point, respectively. It is the correction value for the orientation angle of the measuring station. =206265, , , This is the constant term in the observation error equation; , , , , The formulas are as follows:

[0064] (3)

[0065] (4)

[0066] (5)

[0067] (6)

[0068] (7)

[0069] in, This is an approximate value for the orientation angle of the survey station.

[0070] Secondly, the predicted initial approximate coordinates are substituted into the normal equation formula (2) to solve for x.

[0071] Finally, the judgment If k=0.1mm; if yes, output x, calculate the coordinates of each monitoring point based on the correction x, substitute x into formula (1) to obtain the observation correction V, and then obtain the adjustment values ​​of the observed horizontal angle, vertical angle and slope distance; if no, such as Figure 2 As shown, perform iterative solutions until... .

[0072] S6: To ensure the reliability of the adjustment results, the average gap method is used to search for unstable benchmark points; unstable points are eliminated and the adjustment is repeated; if it still does not meet the requirements, the measurement is repeated and steps S4-S6 are repeated.

[0073] S7: The adjusted coordinates of each monitoring point are organized and transmitted to the corresponding database. These updated coordinate data are then used to optimize and update the rolling prediction model parameters for the next cycle, forming a self-learning closed loop to continuously improve prediction accuracy and measurement stability.

[0074] S8: Using the results obtained from the adjustment in step S5: horizontal angle, vertical angle, and slope distance, update the learning data in the learning file for measurement in the next cycle.

[0075] S9: Repeat steps S4-S8.

[0076] Example 2: Based on the same inventive concept as the total station self-learning measurement adjustment method based on rolling prediction in Example 1, this application also provides a total station self-learning measurement adjustment system based on rolling prediction, such as... Figure 3 As shown, the total station self-learning measurement adjustment system based on rolling prediction includes the following steps:

[0077] Step 1: Deploy the gateway device. Configure the basic information of the gateway and preset the timed task measurement plan within the gateway through the PC / Android terminal remote management platform.

[0078] Step 2: According to the scheduled measurement plan, the gateway sends measurement instructions to the total station; based on the learning document, the total station automatically collects data from the monitoring points, transmits the data back to the gateway via the communication protocol, and performs data adjustment calculations.

[0079] Step 3: The coordinates of the monitoring points obtained from the adjustment are uploaded to the PC / Android project management platform via an HTTP POST request, and the management platform performs data visualization analysis.

[0080] Step 4: The gateway control module will synchronously record the system operation and operation logs throughout the process, and back up the measurement and adjustment results.

[0081] Furthermore, step 2 employs a total station self-learning measurement adjustment method based on rolling prediction as described in Example 1.

[0082] Finally, it should be noted that the above embodiments are not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the technical solution of the present invention are also within the protection scope of the present invention.

Claims

1. A total station self-learning measurement adjustment method based on rolling prediction, characterized in that, The method comprises the following steps: S1: controlling the total station to automatically measure a plurality of monitoring points according to a preset measurement period, and obtaining angle observation values and distance observation values of the monitoring points; S2: calculating measurement errors after each half measurement cycle, and re-measuring the measurement cycle if the error is out of limit; S3: constructing a time series prediction model based on a historical monitoring coordinate database, predicting displacement amounts of the monitoring points in the current measurement period, and obtaining predicted coordinates of the monitoring points; S4: taking the predicted coordinates as initial approximate values, performing three-dimensional adjustment calculation on the angle observation values and the distance observation values, obtaining adjustment coordinates of the monitoring points, and obtaining adjustment values of observation data; S5: searching for unstable reference points by using an average gap method, excluding unstable points, and re-adjusting; if the adjustment still does not meet the requirements, re-measuring, and repeating steps S1-S5; S6: writing the adjustment coordinates in step S4 into the historical monitoring coordinate database, updating a time series prediction model of the next measurement period, and realizing rolling prediction and self-learning update of the monitoring point coordinates; S7: using the observation data adjustment values obtained in step S4 to update learning data in a learning file.

2. The self-learning measurement adjustment method of a total station based on rolling prediction according to claim 1, characterized in that, The time series prediction model in step S3 is a neural network prediction model or a statistical prediction model constructed based on historical monitoring coordinate data.

3. The self-learning measurement adjustment method of a total station based on rolling prediction according to claim 2, characterized in that, The neural network prediction model is a Transformer time series prediction model based on an attention mechanism.

4. The self-learning measurement adjustment method of a total station based on rolling prediction according to claim 1, characterized in that, In step S4, the three-dimensional adjustment calculation adopts a three-dimensional indirect adjustment method to jointly solve station point coordinates, monitoring point coordinates, and observation angles and distances.

5. The self-learning measurement adjustment method of a total station based on rolling prediction according to claim 1, characterized in that, In step S6, the adjustment coordinates obtained in the current measurement period are used to update the historical monitoring database, and the time series prediction model is parameter-modified based on the updated historical monitoring database, so as to realize rolling update and self-learning optimization of the monitoring point coordinate prediction model.

6. A total station self-learning measurement adjustment system based on rolling prediction, characterized in that, The system comprises a gateway control module, a data acquisition and processing module, and a platform management module; the gateway control module is connected with the data acquisition and processing module through an RS485 serial port, and the gateway control module is connected with the PC end / Android end platform management module through an HTTP POST request / MQTT protocol to establish network communication connection; wherein: The gateway control module undertakes the responsibilities of field device control and data transfer storage; The data acquisition and processing module is used to execute the total station self-learning measurement adjustment method based on rolling prediction in claim 1, and complete acquisition, processing and learning file self-update of measurement data; The platform management module realizes device remote management and data visualization based on transmission data.

7. The self-learning measurement adjustment system based on rolling prediction for total station according to claim 6, wherein, The gateway control module comprises three sub-function modules: The device control module is responsible for turning on / off the total station, tilt inspection, triggering measurement, and scheduling of a timing measurement task; The data forwarding module receives original measurement data returned by the total station, transmits the to-be-processed data to the data acquisition and processing module, and forwards the processed monitoring results to the platform management module; The local storage module locally stores the original measurement data and the processed monitoring point coordinate data, and provides support for data tracing.

8. The self-learning measurement adjustment system based on rolling prediction for total station according to claim 6, wherein, The platform management module realizes the following functions through an online management interface: Data visualization: visual display and analysis of monitoring point coordinates and measurement result data; Remote management of equipment: remote monitoring and configuration of the running state of the gateway control module and the total station.

9. The self-learning measurement adjustment system based on rolling prediction for total station according to claim 6, wherein, The data acquisition and processing module includes three sub-functions of data acquisition, adjustment calculation and learning file updating, adopts the total station self-learning measurement adjustment method based on rolling prediction in claim 1, and completes acquisition, adjustment calculation and processing of original measurement data, and self-updating of the learning file.

Images