Comprehensive evaluation of loess collapsibility of oil and gas pipeline based on cloud theory

The comprehensive evaluation of pipeline loess collapsibility risk is a necessary means to control the safety risks of pipelines in the collapsible loess section. It is also one of the critical scientific bases for risk prevention, control, and management. The comprehensive evaluation system of cloud theory consists of quantitative and qualitative indexes, and the evaluation system has the characteristics of randomness and fuzziness. In view of this problem, the standard qualitative and semi-quantitative evaluation methods have intense subjectivity in dealing with the uncertainty problems such as randomness and fuzziness of the system, the cloud theory, which can effectively reflect the randomness and fuzziness of things at the same time, is introduced. The state scale cloud and index importance weight cloud of pipeline loess collapse risk are constructed by the golden section method. The uncertainty cloud reasoning process of the quantitative indexes and the expert scoring method of the qualitative indexes are proposed. The comprehensive evaluation model of loess collapsibility risk of oil and gas pipeline is established, and the engineering example is analyzed. The complete evaluation results of 10 samples to be evaluated are consistent with the results of the semi-quantitative method and are compatible with the actual situation. The evaluation process softens the subjective division of index boundary, simplifies the preprocessing of index data, realizes the organic integration of quantitative and qualitative decisions, and improves the accuracy, rationality, and visualization of the results.

www.nature.com/scientificreports/ on fuzzy membership inference based on the complexity and comprehensibility of the evaluation system. The risk assessment method provides a good idea for the quantitative assessment of the complex system of pipeline geological hazard risk. However, there is still intense subjectivity in the division of the evaluation target and index scale, and the randomness of the system is ignored. There are also cumbersome and time-consuming procedures in the operation process, such as the unified processing of the index dimension. Wang et al. 12 developed a landslide risk assessment system for the long-distance pipeline based on the expert system platform (ES) by analyzing a large number of landslide cases in China's West-to-East Gas Pipeline. This method relies too much on the existing knowledge base and experience judgement. Its application scope and objects are generally strictly limited, and the accuracy of the assessment results is not high. In addition, the late knowledge updating and maintenance of the system platform are of significant workload and high cost. Vasseghi et al. 13 used the finite element analysis (FEA) method to evaluate the response relationship of pipelines to ground displacement in the potential landslide area, and combined with theoretical analysis, revealed that the main reason for bending failure of pipelines was the high shear stress at the girth weld. Although numerical simulation analysis methods have high accuracy, its simulation boundary assumptions for monomer disasters complex variability among different location, geological, environmental conditions, and simulate the required parameters in general still need using soil mechanics test. Therefore this method does not apply to the universal development of the pipeline in loess collapsibility disaster danger evaluation and sorting work quickly. Teng et al. 14 used earthquake scenario analysis and spatial network method to classify and evaluate the seismic damage of urban underground pipelines. It proposed a pipeline disaster impact assessment model based on a geographic information platform. As a result, the form and results can be better grasped. It also predicts the susceptibility of geological disasters along the route to the division and segmentation. It visually displays the relevant developments in the form of maps, but it is not suitable for evaluating the risk status of individual geological disasters. In addition, the researches on loess subsidence disasters of oil and gas pipelines in existing works of literature primarily focus on the analysis of the influencing factors and development mechanism of the classification of uneven settlement types in loess caves and gullies along the pipelines in the loess region 5,[15][16][17][18][19] . There are also studies on mechanical behaviors and monitoring and early warning of stress-strain subsidence and suspension of pipelines based on the tubing-soil coupling model or elastoplastic foundation model [19][20][21][22][23][24] . Yu et al. 25 evaluated the hazard of loess collapse using the fuzzy mathematics method based on the investigation of pipeline geological hazards. Gao et al. 26 based on the principle of coupling synergy (CCP), evaluated the pipeline loess collapsibility risk from two subsystems of soil and pipe body and provided better guidance for engineering practice. Yang 27 and Zhou 28 established a fuzzy comprehensive evaluation model based on field investigation and hazard evaluation analysis of pipeline loess collapse disasters and applied it to different pipe sections in loess areas. The overall effect was better than the subjective qualitative evaluation results. On the other hand, the above-related studies [15][16][17][18][19][20][21][22][23][24][25][26][27][28] , on the one hand, paid little attention to the risk analysis and evaluation of the pipeline loess collapsibility disaster with universal development. The disaster-bearing environment's comprehensive factor analysis of the disaster-bearing capacity of the disaster-causing factors was not systematic and in-depth. In addition, the selection of indicators was not extensive. On the other hand, index data acquisition and processing showed more complicated time-consuming and continuity of quantitative and qualitative discrete data fusion. Participate in the evaluation process of subjectivity is strong, the fuzziness and randomness model for the complex system in operation to consider problems of one sector. Therefore, it is necessary to continuously optimize and introduce new ideas and propose new methods based on previous studies to provide diversified and high-quality scientific and technological support and services for the risk assessment of pipeline loess collapsibility disasters. Pipeline loess collapsibility risk is a system composed of loess collapsibility in laying area and pipeline vulnerability, affected by many factors 15,29 . Its evaluation organization is characterized by multiple attributes, multiple levels, and multiple indicators and has both quantitative and qualitative indicators. It is the key to determining the success or failure of the evaluation and the accuracy of integrating these uncertain indicators, which have randomness, fuzziness, and different magnitude and dimension, and participate in the risk evaluation. Cloud theory 30 is often used to describe the relationships between qualitative and quantitative uncertainty things into a powerful mathematical tool, integrates the advantage of fuzzy mathematics and probability statistics, can effectively reflect the properties of things at the same time the randomness and fuzziness, and constitute a mapping between qualitative and quantitative, makes the qualitative concept and quantitative numerical transformation between clear, specific, and controlled. In the past 20 years, this theory and method have been widely used in system evaluation and control, data mining, group decision making, and other fields [31][32][33][34][35][36] . This paper takes the loess collapse of oil and gas pipelines as the research object. It proposes a comprehensive evaluation model of loess collapsibility based on standard cloud theory, which simplifies the preprocessing process of quantitative data and optimizes the cloud method and qualitative evaluation results. The results of cloud theory are compared with those of extension theory, PGRMS, and expert evaluation. The validity and rationality of the model are verified, and it provides a reference for the study of geological hazards in oil and gas pipelines.

Methodology
Construction of comprehensive evaluation system for pipeline loess collapsibility risk. The loess collapse disaster of oil and gas pipelines is the loess collapse of pipelines (including historical and potential) as the disaster-causing body, and the disaster-causing body of the geological disaster system, which takes the pipeline engineering body as the carrier, is under certain engineering geological conditions (topography, landform, tectonic activity, hydrogeology, stratigraphic lithology, etc.) and induced conditions (irrigation and precipitation). The disaster-bearing body is the main body of the pipeline project, including steel pipelines and anti-corrosion coatings, station valve chambers, etc., the laying of the channels (position, method, burial depth, etc.) in the loess-affected range, and the improvement of pipe trench soil. Filling, tamping, hydraulic protec-tion, etc., directly affect the carrying method and carrying capacity of the pipeline. Besides, these complex factors together constitute the vulnerability factor of the risk assessment. And a series of prevention and control measures in pipeline maintenance (excavation, backfilling, tamping and filling hydraulic change measures, etc.) together constitute the vulnerability factors of risk assessment.
Based on the analysis of the constituent elements of the pipeline mentioned above, loess collapsibility hazard system, combined with related research results 5,[15][16][17][18][25][26][27][28][29] , the independent representative adaptability of each indicator and the hierarchy of the evaluation system are considered and optimized and proposed a multi-level comprehensive evaluation index system consisting of four first-level indicators and sixteen second-level indicators. As shown in Fig. 1, there are eight quantitative indicators and eight qualitative indicators among the sixteen second-level indicators. The dimensions, level, attribute, etc., are different. How to organically integrate these complex indicators to participate in model calculations is the focus of evaluation research.
Cloud theory. In a certain amount of universe U, there is X ⊆ U, and T is a qualitative concept in U space.
When the value x (x ∈ X) is T on the membership C T (x) ∈ [0,1], it is stable A random number with a tendency (such as a normal distribution), that is, there is formula (1), then the mapping of the concept T from the universe U to the interval [0,1] in the number domain space is called cloud 30 , and each x is a cloud drop; the digital feature C (Ex, En, He) 30 is the control parameter of cloud distribution, where Ex is the central value of the universe, called expectation; En is the range of cloud droplet distribution that can be accepted by qualitative concepts in the universe space (that is, the degree of fuzzy picture), called entropy; He is the entropy of En, which is jointly determined by its randomness and ambiguity (that is, the degree of condensation of cloud drops in the universe of discourse, as shown in Fig. 3), and is called hyper-entropy. This paper introduces a one-dimensional regular cloud generator with universal applicability 32 , and its main algorithm is as follows [30][31][32][33][34][35][36] . (1) generating a normal random number taking En as expectation and He as standard deviation, wherein norm is a function generating random numbers which obey normal distribution; y i = norm En, He 2 (2) generating a normal random number with Ex as an expected value and Y 2 as a variance; x i = norm Ex, y 2 (3) Calculating the certainty of random number (4) output a cloud droplet; (x i , µ i ) And (5) repeating the steps (1) to (4) until the required N cloud droplets are generated, and the N cloud droplets form a cloud to realize the uncertainty cloudification of the concept. (1) the numerical characteristics (Ex, En, He) and the quantized value x of the concept are known, and a normal random number En' is generated by taking En as an expected value and He as a mean square deviation; (2) Calculate the degree of certainty y that the quantitative value X belongs to a certain concept, namely: Y-conditional cloud algorithm.
(1) Given the numerical characteristics (Ex, En, He) of the concept and the certainty y, y ∈ [0,1], a normal random number En′ is generated with En as the expected value and He as the mean square deviation; (2) calculating a quantitative value X satisfying the certainty y, namely: Hybrid cloud algorithm. For two clouds C1 (Ex1, En1, He1) and C2 (Ex1, En1, He1) in the same universe, the mixed operation relation is (8)- (12), which can also be extended to n clouds. www.nature.com/scientificreports/ Comprehensive evaluation procedure for pipeline loess collapsibility risk. If the evaluation set layer (measured value + expert score value) in Fig. 1 can be mapped into a cloud with a specific distribution range and law, that is, a standard distribution cloud. In addition, it is assumed that the determination degree of each evaluation state in the common set layer is also subject to cloud distribution. Then, the comprehensive evaluation process of loess collapsibility risk of oil and gas pipelines based on cloud theory is simplified, as shown in Fig. 2. The specific steps are described as follows: related derivation and operation in the evaluation process and the driving of cloud generator are realized using MATLAB software.
Step 1 Determination of risk state target cloud and evaluation index scale interval.
To correspond to the results of the recommended method 2 in the pipeline geological disaster risk management norms, this paper adopts 5-grade evaluations of high, relatively high, medium, relatively low, and low to scale the status of pipeline loess collapsibility risk as shown in Fig. 1 standard set. The golden ratio method 37 is used to divide the language into five levels in the domain [0,1]. The cloud number characteristics and distribution of languages in each class are shown in Table 1 and Fig. 3. Similarly, the results of the 5-level evaluation interval division of 16 s-level indicators in the evaluation set are listed in Table 2. The eight quantitative indicators in Table 2 are divided into bilateral constrained approximations 32 . The eight qualitative indicators were scored and graded by experts on site 2,7,9,38 , and then clustered according to the cloud parameters of the corresponding level in Table 1.
Step 2 Evaluate the indicators and the cloud transformation of samples to be evaluated. For the eight quantitative indicators U 11 , U 12 , U 22 , U 24 , U 31 , U 33 , U 34 , and U 43 in Table 3, there are comment levels with upper and relatively low boundary values of x 1 ij and x 2 ij , and their normal cloud numerical characteristics can be expressed as: The five-grade scale interval of each evaluation index was determined Step 1 Determination of risk assessment standard set Step 2 Evaluation indicators and sample cloud transformation Cloud transformation of quantitative measured data(C Quantitative evaluation ) Uncertain reasoning Cloud transformation of qualitative scoring data (C Qualitative evaluation ) Uncertain reasoning Step 3 The evaluation index system gives weigh The Golden Section Ratio Driven Five-stage Scaling Method The expert group determines cloud weights C The weight Step 4 Evaluation results cloud computing and analysis Cloud hybrid computing TC The results

Uncertainty passes and jumps
Visual comparison of similarity Proximity analysis  www.nature.com/scientificreports/ Since the boundary value refers to the existing research [6][7][8][12][13][14][15][16][17][18][19]26,31 and expert suggestions, the indicators of continuous quantity description are subjectively divided according to specific standards and experience, but they Figure 3. Cloud chart of evaluation criterion and weight. From the cloud atomization ambiguity CD = 3He ij /En, He ij generally takes a number between 0 and En/3 to satisfy the Gaussian cloud distribution 19 . Through many experiments in this paper, the practical value that is convenient for calculation is preferred. Generally, the larger the He ij , the thicker the normal cloud, and vice versa (Fig. 3).
Determine the level standardized cloud parameters of 8 quantitative indicators through formulas ⑿, ⒀, and multiple experiments, as shown in Table 3, and then build a single-condition single-rule cloud generator 23,24 to realize the cloud conversion process of the measured value x A of the quantitative index. The single-condition single-rule cloud generator is shown in Fig. 4. According to formulas ⑹ and ⑺, firstly generate a normal random number En A ′ with En A as the expected value and He A as the mean square error, and calculate the certainty y = exp[-(x A -Ex A )2/2(En A ')2]; Regenerate the normal random number En B ′ with En B as the expected value and He B as the mean square error. If the antecedent activates the rising or falling edge x A ≤ Ex A or x A > Ex A , the latter is the same. Activate the value of rising or falling edge x B = Ex B -En B ′ × (− 2lny)0.5 or x B = Ex B + En B ′ × (− 2lny) 0.5 .
The output results of the X-condition cloud and Y-condition cloud in the regular cloud generator are obtained through random processes. Therefore, for the same input, the output value received by the cloud inference method each time has uncertain characteristics. It fluctuates up and down within a reasonable range and has a stable tendency as a whole, obeying Gaussian distribution [21][22][23][24] . Through the cloudification of quantitative indicators: On the one hand, the subjective division of the level boundary is softened; on the other hand, x A of different dimensions and magnitudes are calculated by X, Y-conditional cloud generators, and the data is dimensionless and normalized. This simplifies the pre-processing procedure. This method ensures the effective transmission  www.nature.com/scientificreports/ and inheritance of uncertainty in the evaluation process, which is significantly better than other quantitative evaluation theories. For the eight qualitative indicators of U 13 , U 14 , U 21 , U 23 , U 32 , U 41 , U 42 , and U 44 in Table 3, it is impossible to use quantitative values to characterize the attributes accurately. This article is based on the scores of the expert group (10) members, whichever is Average 25 , score and evaluate eight qualitative indicators one by one; after obtaining the corresponding values, process them according to the process of cloudification of quantitative indicators and participate in the evaluation system.

Evaluation results integrated cloud computing.
Step 3 Determination of the weight cloud of indicators at all levels of the evaluation system.
The weight reflects the relative importance and contribution rate of the various factors in the evaluation system to the goal and the key "hub" for the continuous transmission and rise of cloud theory. The risk evaluation indicators of loess collapsibility in oil and gas pipelines are primarily qualitative. It is difficult to directly obtain the weight value of each hand through objective weighting processes such as entropy method, coefficient of variation method, and gray correlation method; at the same time, to avoid analytic hierarchy process, etc. The shortcomings of personal empowerment are artificially arbitrary. This article invites seven industry experts and three technical backbones of pipeline operators to form an expert group to carry out the weighting of the evaluation system and the scoring of sample qualitative indicators. The language description of the relative importance of the participating indicators is not essential, second important, general, and more critical. And the significant 5-level state, based on the golden ratio-driven method, the index weights are assigned 5-level scale weights on the domain [0,1]. The corresponding cloud digital features and clouds are shown in Table 1 and Fig. 3, consistent with People's scientific understanding of things. Thus, while eliminating subjective arbitrariness, transforms qualitative language into weight values for calculation, reflecting the ambiguity, randomness, and complexity of the evaluation process.
Step 4 Integrate the evaluation results into cloud computing. Following the evaluation index system and cloud weight determination process of cloud conversion in the previous article, according to the hybrid cloud calculation formulas (8)-(11), the weighted average method is used to transfer the evaluation results of the bottom layer of the comment concentration to the upper layer, and then to the target layer, which is the risk of the evaluation object. The resulting cloud can be used for visual analysis and decision-making with the ruler cloud.
In the formula, TC is the comprehensive cloud of the evaluation target, C i and ω i are the i-th index cloud and the weight cloud, respectively, and ω i C i represents the multiplication of the two clouds. This process transfers uncertain reasoning and enables the concept of cloud theory to rise continuously.

Analysis of engineering examples
Nearly 40% of the length of the Lan(zhou)-Cheng(du) Crude Oil Pipeline, a representative project of China's north-south oil transport, is laid in the Longxi Basin of the Loess Plateau, which is a typical collapsible loess distribution area. This area is located in the southwestern part of the Loess Plateau, bordered by the Qinba Mountains in the south and Wushaoling in the west. It is a continental temperate semi-arid climate zone with significant temperature differences and slight rainfall. The annual average temperature ranges from 4 to 14 ℃. Precipitation accounts for 50 to 70% of the yearly total, which is also a period when geological disasters frequently occur. Most of the land types in the area across which the pipeline is laid are agricultural land, and the primary vegetation is the cash crops. The vegetation coverage rate is generally at a medium level. Data sources. According to the 2020 pipeline geological disaster survey data, there are 109 geological disasters of various types in the Longxi Basin section of the Lancheng Pipeline, and there is 54 loess collapses, accounting for 49.54% of the total disasters. It can be seen that the loess collapses are densely developed and are prone to The characteristics of Gaofa. This paper randomly selects ten samples of risk assessment from 54 pipeline loess collapse disasters, and their specific locations are the pipeline mileage K7 + 110, K27 + 780, K32 + 720, K43 + 460, K43 + 920, K47 + 940, K57 + 900, K87 + 440, K168 + 850, K257 + 733 (numbered as LC 1-LC 10).
The essential characteristics of the selected loess collapsibility disasters are shown in Table 3. The initial data of the eight quantitative indicators of 10 samples to be evaluated are obtained from the existing meteorological, geographic, and geological data and on-site measurement. The results are listed in Table 4; the qualitative indicators are invited to the site according to the third step of the methodology. Judgement, record each expert's scoring value for different hands of each loess collapsibility and find the average value as the participating data 7,9,38 . Table 5 only lists the scores of 10 experts on the collapse disaster activity U14 in the sample indicators to be evaluated to simplify the length. The parameters of the remaining seven indicators can be obtained in the same way and participate in the evaluation system.

Calculation of results.
First, use the single condition single rule cloud inference (1000 cycles of MATLAB generator) algorithm in the second step to calculate the cloud digital feature values of the 16 indicators in the 1-10 loess collapsibility, as shown in Table 6 Column. Then use the algorithm in formulas (8)- (11) to compare the index cloud C evaluation and cloud C weight in Table 6, according to the fourth step Chinese formula (14) upload the second-level index U ij to the first-level index U i and then pass it to the target layer U of the evaluation set. Afterward, according to steps 1-5 of the forward cloud algorithm, the TC results were generated to generate the pipeline loess collapsibility danger cloud as shown in Fig. 5 superimposed the predetermined ruler cloud, and visualized comparison and closeness analysis were performed to determine the sample risk status level. Figure 5 shows that the risk result cloud of loess collapsibility of No. 1-10 which are basically similar to the ruler cloud in the form of the "skeleton", are distributed between the "relatively low" and "relatively high" cloud rulers, and mainly concentrated in Between "medium" and "relatively high" cloud ruler. The specific manifestations are as follows: LC1 cloud is inclined to "relatively high cloud"; LC2 cloud is close to "relatively high cloud"; LC3 cloud is inclined to " medium cloud"; LC4 cloud is close to " medium cloud" and LC5 cloud is near " medium cloud"; LC6 cloud is adjacent to "relatively high cloud"; LC7 cloud is adjacent to " medium cloud"; LC8 is close to "relatively high cloud"; LC9 cloud is close to " medium cloud"; LC10 cloud is close to "medium cloud". The hazard degree of loess collapsibility to oil and gas pipelines is in the order of LC6 > LC2 > LC8 > LC1 > LC4 > LC 2 > LC3 > LC5 > LC9 > LC10.

Result analysis.
To test the applicability and effectiveness of cloud theory in the evaluation of loess collapsibility risk of oil and gas pipelines, the extension theory method (a method based on fuzzy mathematics) 39-41 , PGRMS method 38 (a semi-quantitative way recommended by the code 2 , which is automatically realized by the pipeline geological disaster risk management system provided by China Petroleum Pipeline Science and Technology Research Center) and expert qualitative assessment method are used to carry out a risk assessment on ten loess subsidence sites respectively. The results are listed in Table 7.
Combining Table 7 and Fig. 5, we can see that the risk of loess collapsibility of No. 1-10 using the PGRMS method and extension theory method is consistent with the results of the method proposed in this paper, indicating the risk of loess collapsibility of oil and gas pipelines based on cloud theory The comprehensive evaluation method is practical and feasible, and can fully reflect the fuzziness and randomness of the participating variables in the evaluation process, which is more in line with people's thinking mode and expression habits on the understanding of the dangerous development of pipeline loess collapsibility, and also enables the evaluation results The expression of is visualized in the form of cloud drop distribution. www.nature.com/scientificreports/ The results of this method were compared with those of the other three methods. The results showed that: (1) The consistency ratio of the evaluation results between the cloud theory method and the PGRMS method was 70%, the relatively high proportion was 20%, and the relatively low proportion was 10%. (2) The consistency ratio of the evaluation results between the cloud theory method and the extension theory method is 60%, the high proportion is 20%, and the low proportion is 20%. (3) The consistency ratio of cloud theory method and expert qualitative method is 30%, the high proportion is 50%, the low proportion is 20%; (4) The consistency ratio between PGRMS and expert qualitative method is 40%, with high proportion 40% and low proportion 20%. (5) The consistency ratio between PGRMS method and the extension theory method is 40%, 50% relatively high, and 10% relatively low. (6) The results of the extension theory method and the expert qualitative method are consistent with the proportion of 40%, the high proportion of 40%, the low proportion of 20%.
In summary, it is found that the cloud theory method introduced in this paper has a relatively high overall risk level for the pipeline loess collapsibility risk assessment result than the other three methods, which is more consistent with the actual situation of the example samples and is more conducive to the safety guarantee of pipeline operators in the study area. The evaluation results of the cloud theory method have high consistency with those of the PGRMS method and extension theory method but poor consistency with the evaluation results of the expert qualitative method. This is mainly because the cloud theory method, PGRMS method, and extension theory method all carry out quantitative processing on comprehensive indicators and carry out fusion operation through a particular mathematical model, which weakens subjective arbitrariness to a certain extent, and the factors involved in the evaluation are more comprehensive, and the evaluation process is systematic. However, the evaluation results of the expert qualitative method only rely on the macroscopic grading judgment of the expert group according to the individual experience. In the process, the subjective arbitrariness is strong. It is easy to be interfered with by different factors such as personal preference, professional background, comprehensiveness, and degree of understanding of various experts. As a result, the accuracy of the results is generally poor.

Conclusion
(1) The risk evaluation system of oil and gas pipeline loess collapsibility comprises a multi-level multi-objective complex index system. Common evaluation theories and methods cannot deal with the ambiguity and randomness of the system at the same time. Based on the analysis of the fuzziness, randomness, dimension, and magnitude difference of pipeline loess collapsibility evaluation indicators, this paper proposes a comprehensive evaluation model of the single loess collapsibility risk of oil and gas pipelines cloud theory. Through the risk assessment and verification analysis of 10 loess collapsible disaster points in the Lancheng Crude Oil Pipeline, its practicability and effectiveness are shown.
(2) The quantitative data in the evaluation comes from actual measurement and background data, and the qualitative information comes from the scoring and assignment of the expert group. The initial data source is relatively objective and reliable; the index weight cloud is determined by the 5-level golden ratio method, which is transmitted. The effective weighting of quantitative and qualitative indicators is realized based on system fuzziness and randomness. www.nature.com/scientificreports/ (3) The single-condition-single-rule cloud conversion method is adopted for the quantitative indicators in the evaluation, which effectively transmits and inherits the ambiguity and randomness of the system, weakens the subjective division of hierarchical boundaries, and simplifies the pre-processing of initial data; for qualitative indicators After scoring and averaging the expert group, the single-condition-single-rule cloud conversion method is adopted, and the interference of subjective factors can be reduced to a certain extent by weakening the division of the grade boundary. (4) The evaluation result is a cloud composed of three parameters: expected value, entropy, and super-entropy, which realizes the evaluation process of quantitative and qualitative integration, integrated decision-making, and the visualization of the final result is a complex system and is used to prevent the loess collapsibility disaster of oil and gas pipelines. Governance provides scientific and technological support and also proposes an effective new method for related research.