Model construction needs to formulate its construction process so that the model can be completed in an orderly manner. Through the practice of software development, we have made a model making process:
(1) data collection, collation and inspection;
(2) Import the data into the modeling system software to establish the model of a single basin;
(3) According to the imported data, the corresponding relationship is sorted out, and geological bodies such as faults, strata and lenses are established in turn;
(4) Check and verify the established three-dimensional visualization model of sub-basins, and if there are any errors, repeat the above steps to modify the model;
(5) Combine the sub-basin models into a complete three-dimensional visualization model.
(2) Scale design of three-dimensional geological model of groundwater in Heihe River Basin.
The modeling area of three-dimensional groundwater model has the characteristics of wide horizontal area and shallow vertical depth, so it is necessary to set different horizontal and vertical scales, and the comparison of horizontal and vertical scale values should be appropriate, so as to establish the model and have a beautiful appearance.
The work area in Heihe River Basin is about seven or eight hundred kilometers long and 400 kilometers wide, while the depth of underground exploration is only a few hundred meters or even dozens of meters. If the model is built in a uniform scale, then the model will be reduced to a face. The geological structure of the underground can't be expressed and can't be browsed. If the vertical scale is enlarged, the Qilian Mountain in the basin is more than 5000 meters above sea level, which is dozens of times lower than the depth of underground exploration data. The height of the completed model is high, and the underground part will be very thin.
In this case, it is necessary to determine the appropriate length-width ratio to make the model more beautiful and meet the actual requirements. The data input of the modeling system is unified in meters. After many experiments, we have determined that the vertical and horizontal scale of Heihe River Basin simulation is 2000. That is to say, the XY direction is reduced by 2000 times when input, while the length of longitudinal data (such as section depth) remains unchanged when input. After such modification, the length-width ratio of the model is appropriate, which can better reflect the geological structure of Heihe River Basin.
(3) The design basis of module modeling and the determination of the boundaries of each module.
The geological structure of Heihe River is complex, and many basin units are developed under different tectonic backgrounds, which leads to the independence of the basin, and also determines the spatial change of groundwater system structure and the complexity of groundwater movement.
To build a geological model of such a large area in Heihe River Basin, the original data read by the system can reach hundreds of megabytes, and the memory occupied by the number of triangular meshes generated by strata will require a lot of machine resources. However, our software positioning is to use microcomputer instead of workstation modeling, so we must consider solving this contradiction.
According to the geological structure, the basin can be divided into several basin units, and each basin unit can be modeled independently. This helps us to solve the problem, that is, to build sub-models of each basin in a unified framework, and then assemble the triangular mesh surfaces of each sub-model to form a total basin model. It not only overcomes the limitation of machine speed, but also ensures the integrity of the whole model. It can be said that it is multifaceted.
The whole model construction process is an integration-separation-integration process. Firstly, the overall framework is established, that is, the fault and surface models of the whole work area are established, then the sub-models are constructed strictly according to the coordinates according to the geological conditions, and finally assembled under the same surface according to the coordinates.
The division of basin units in the basin is based on geological structural characteristics. Faults and uplifts play a great role in the migration and change of groundwater, and are the natural boundaries of the basin. According to geological data, the dividing line can be marked on the basin map. Then the geological data are divided into regions according to the dividing line, and then the model is established according to the geological data of each region. Seven independent models are Damaying Basin, Shandan Basin and Daqingyang Basin, Zhangye Basin, Jiuquan East Basin, Jiuquan West Basin, Jinta Basin and Ejinaqi Basin. Specifically, the location of the dividing line of each basin is described as follows:
(1) Zongzhai-Gu Yong fault is the dividing line between Zhangye basin and Damaying basin;
(2) Gu Yong uplift is the dividing line between Zhangye Basin and Shandan Basin;
(3) The arc zone between Kushuishan Uplift and Gao Tai Concealed Uplift at the western end of Yumushan Mountain constitutes the dividing line between Jiuquan East Basin and Zhangye Basin;
(4) The faults on the north and south sides of the Central Mountain north of Jiayuguan are the dividing line between Jiuquan Basin and Jinta Basin;
(5) Jiayuguan fault is the dividing line between Jiuquan East Basin and Jiuquan West Basin;
(6) Diwan Liang Dong Uplift is the dividing line between Jinta Basin and Ejinaqi Basin.
(4) Construction and design of sub-model.
For the sub-model, that is, a specific basin, the geological structure is relatively complex, and there are many geological bodies to be represented, and there is a cutting and intersecting relationship between these geological bodies, which requires the construction of these geological bodies one by one in a certain order. Here is a brief description of the order of model construction, that is, reading in the original data to generate the surface, establishing the fault model according to the fault line of the surface and the fault line on the selected section, then constructing each stratum, specifying the effective area of the stratum, constructing the lens, adjusting the spatial position between the stratum and the lens, and finally generating the geological body. After the geological body is generated, visual operation and graphic output can be carried out.
In order to explain modeling in this order, it involves the relationship between modeled geological bodies. Generally speaking, all data need a certain geological body to play a reference role in the modeling process, and the surface can play this role well. After inputting the local surface, fault lines and section lines can be drawn on the surface, and they fluctuate according to the fluctuation of the surface. Therefore, it is necessary to build a model of the surface first. After the local table is established, the metadata of watershed boundary map and surface geographic information map can be input on the surface. When constructing strata and fault geological bodies, the footwall and footwall of the fault are formed because the ground plane is cut by the fault plane, and it is not necessary to change the fault plane in this process. Therefore, it is necessary to establish the fault plane first, and then the site plane, in order to automatically cut the ground plane. After the horizon was built, the camera followed. Because the lens belongs to different strata, it needs to be built on the ground level. After the lens body is generated, so far, all preparations have been completed and geological blocks can be generated. This process is an optimization process of establishing the model, which has been proved to be correct in practice.
(5) Refinement and elimination of surface data points
The surface data is converted from ARCINFO format data, and the plane line segments that make up the contour line correspond to their elevation values one by one, so there is a three-dimensional coordinate set of the surface. When the system inputs, the points that make up the isoline are read in, that is, the points of each line segment are read in, and the discrete points needed to generate the model are formed by combining the elevation values corresponding to the line segments. The system uses these discrete points to interpolate to generate the ground. Due to the large amount of data in Heihe river basin 1∶250000 and the high density of discrete points, the data stored in MAPGIS plain file reaches more than 150 trillion. If all these data are input into the system for interpolation to generate the surface, then the number of small triangles generated on the surface will be quite large, occupying too much computer resources, which will have a great impact on the subsequent model construction. At the same time, the speed of generating the surface is slow and the effect is not ideal. So we should dilute these discrete points. The effect of thinning is to reduce the number of points on the isoline. Because there are a lot of points on the isoline, scaling will not affect its accuracy. After refinement, the density of discrete points needed to generate the model is reduced, the number of discrete points is reduced, the speed of generating the surface is accelerated, and the smoothness of the surface is improved.
For some bad points on the surface, such as points with too high or too low elevation, that is, the elevation is higher than the highest point or lower than the lowest point on the surface, these points are caused by errors or data conversion. Using this point interpolation, the surface will fluctuate violently and the surface will be rough, which will affect the smoothness of the surface. Such points need to be removed when entering the system. That is to say, in the system input module, the threshold is used to limit, and the points that are too high or too low are eliminated and are not allowed to participate in modeling.
(6) Grid size design of three-dimensional geological model of groundwater in Heihe River Basin.
From the modeling principle, we can know that the geological bodies constructed by the model, such as strata and faults, are all connected by grids to form faces and surrounded by faces. The smallest unit of the model is a small triangle, and the number of triangles directly affects the accuracy of the model and the running speed of the system.
Generally speaking, if the triangular mesh of the generated model is too large, the surface of the model is rough and the model is not fine, and even the morphological characteristics of the surface cannot be expressed. If the grid is too small and the grid density is high, the operations on these triangles occupy a lot of system resources, which makes the amount of data processed by the computer increase sharply, thus making the machine run slowly. If the amount of discrete point data is too much or too little, it will make the model surface more complex and unable to express the overall characteristics of the model surface. Therefore, it is necessary to choose the appropriate grid size when establishing the model.
In the process of constructing the basin model of Heihe river basin, through practice, the grid size of the surface is 50 ~ 200 m, the grid spacing is good, and the machine speed and surface smoothness can be coordinated and unified. Generally, in the initial construction of the model, the grid spacing of 200m is selected, and the machine speed is high. When the model of each small watershed is completed, the grid spacing of 50m is selected to generate the surface, so that the surface is fine.
Because the strata and faults are modeled by line segments on the profile, the system will automatically encrypt the discrete points on these line segments and select the grid spacing of 200m, which has little influence on the formation of strata and faults and meets the requirements in fineness. Due to the small area and moderate lens accuracy, the grid size of 100m can meet the needs.
For raster data, such as remote sensing images, the spatial resolution can be adjusted according to the accuracy of model display. General 100 ~ 300 dpi is enough. Fine remote sensing images of more than 800 megabytes can be generated into BMP, JPG and other graphic formats as needed, ensuring pixel accuracy and reducing memory occupation.
(7) Automatic generation and local adjustment of the model
The system discretizes the fault plane, ground plane and the top and bottom of the lens according to the line segments selected by technicians on the section, and then interpolates these discrete points to form a series of small triangles to form the faces of different geological bodies. In this way, the process of generating surface will cause the surface to extend according to the trend of discrete points, which will lead to the deformation and distortion of geological surface, which is inconsistent with the actual geological situation. For example, a series of stratigraphic relationships such as mutual cutting between faults, fault cutting strata, and cross-cutting strata cannot be established.
In this case, it is necessary to use control points to locally fine-tune the face and change the shape of the face. Make the generated surface as close as possible to the interpretation results of geological data and express the real geological situation. At the same time, we can also establish the relationship between geological bodies such as mutual cutting between faults, strata cutting by faults, and cross cutting of strata. Then the system defines the relationship between these surfaces according to technicians, and automatically generates qualified geological bodies.
Therefore, in the process of model construction, it is necessary to use the functions of system automatic generation and local adjustment of control points to complete the model construction.
(8) Error analysis and treatment of three-dimensional geological model of groundwater in Heihe River Basin.
In the process of model construction, it will inevitably lead to the generation, transmission and accumulation of errors. It is necessary to analyze the causes of errors and establish an error control mechanism to improve the quality and accuracy of the model. From the analysis of error sources, there are reasons for basic data and errors generated by the system in the modeling process, which are discussed separately below.
1. Error in basic data
The model construction mainly uses profile data and borehole data, and the borehole data will make the generated stratum fluctuate too much because the borehole stratum changes too disorderly. In the aspect of section drawing, errors will also occur due to the influence of human factors in the drawing process.
Generally speaking, different technicians have different understandings and treatments of geological structures, which will inevitably be reflected in the stratigraphic changes and numerical values of the profile. In addition, using different geological data, some geological data have different arguments and even contradictory views, which leads to errors. In the process of drawing, technicians use different data for synthesis and processing, such as drawing, which makes mistakes happen. When drawing by hand on grid computing paper, especially when drawing long sections, errors will inevitably occur in the drawing process due to the folds of the paper. Even some calculation papers are not standard, and the grid lines are oblique. At the same time, errors will be transmitted in the process of drawing scanning and vectorization. The outstanding performance is that the position of stratigraphic line cannot intersect at the intersection of sections, or the strike of strata is inconsistent with that of other sections.
2. Errors in system operation
Because the basic data of the model will produce certain errors in the data production process, when the data with errors are input into the modeling system, the errors will be transmitted and amplified, which will reduce the accuracy of the model. In the case of avoiding errors in basic data as much as possible, it is also necessary to analyze the errors in system operation.
By analyzing the errors in the process of modeling, it can be concluded that errors may occur in all steps of modeling due to operational errors or the setting of system accuracy. The following analysis can be made specifically. In the data processing stage, there will be errors in the data fusion of various data and graphic materials, such as projection transformation and error correction. In addition, in the interpolation process of system modeling, local points are dense, such as the selected section line, and there are few discrete points in other areas. Interpolation according to mathematical methods will lead to the phenomenon that the interpolated surface does not match the actual surface. In order to make the surface more realistic, the method of pulling or pressing with control points is adopted, which will inevitably cause certain errors.
In addition, when the distance between model layers is small, it is easy to cross and overlap with the ground plane. When using control points for correction, the lens body or ground plane changes, which is easy to produce errors. In addition, when the position between the ground plane and the lens body is adjusted, a chain reaction is easy to occur, leading to errors.
3. Error control
Some mistakes are inevitable and some are accidental. In view of these errors, our main strategy is to check and correct the errors in basic data, such as strengthening quality control, enhancing the sense of responsibility of technicians, allowing geological technicians to coordinate and correct maps in the same area, and hiring geological experts to check and verify. These measures have played a very good role in quality control. For the errors in the system operation, the project team members analyzed the reasons, strengthened debugging and took various measures to minimize the errors. If the profile position is adjusted accurately and the stratum position is consistent as far as possible, for obvious errors, such as stratigraphic line mismatch, the change of stratum is analyzed according to other profiles, and the erroneous stratigraphic line is edited by using the editing function of the system to minimize the error. In the process of generating strata, try to add more points on the profile line, so that the formed stratum grid passes through the profile line, and so on. These measures have played a very good role in reducing errors.
Model checking in modeling is also a way to reduce errors. Model checking, including the accuracy check of the original data, that is, whether the formed surface is consistent with the original data points and whether the data of the original points are retained; At the same time, geological rationality can be tested by using geological data such as drilling holes and profiles. In addition, we can also make a section near the original section by cutting the model, and compare the two sections to see the geological boundary of each layer (including fault plane). Some important sections can be used to check the consistency of formation morphology at the intersection of faults, the intersection of faults and strata, different layers and both sides of faults. The model verification process plays a real-time verification role in the error control of the model.