Method for predicting residual crushing mechanical property of pitted steel pipe based on combination of digital twinning and deep learning
Through the combination of digital twins and deep learning, the problem of predicting the mechanical properties of residual crushing pitted steel pipes is solved, and accurate mechanical performance evaluation is achieved, which reduces maintenance costs and improves structural safety.
Patent Information
- Application Number
- CN202510438267.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-08
- Publication Date
- 2025-08-01
AI Technical Summary
The prior art cannot accurately evaluate the uncertainty of residual strength caused by pitting damage in steel pipe structures, resulting in increased maintenance costs and safety hazards.
Using digital twin technology and deep learning, we use the numerical model of steel pipes containing pitting damage to generate basic data sets, train deep learning models, extract pit features of pitting, match the axial collapse intensity, and achieve accurate prediction.
Accurate mechanical properties prediction of pitting steel pipe structures are achieved, which reduces maintenance costs and improves the safety and reliability of the structure.
Smart Images
Figure CN120409202A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of steel pipe mechanical property prediction, and specifically to a method for predicting the residual crushing mechanical properties of pitted steel pipes based on the combination of digital twins and deep learning. Background Art
[0002] Steel structures are a common structure used extensively in civil engineering and marine engineering. Steel pipes, in particular, are often used for connections and load-bearing structures. During the service life of steel pipe structures, in addition to load-bearing damage, corrosion damage, such as uniform corrosion, pitting corrosion, and erosion corrosion, also has a significant impact on the degradation of their mechanical properties. Among these corrosion conditions, pitting corrosion is the most common form of corrosion on steel pipe surfaces, leading to localized deterioration and premature failure. These randomly occurring pitting corrosion on steel pipe surfaces significantly increases the instability of the structure's mechanical properties, making their prediction difficult and challenging.
[0003] Currently, for steel pipe structures widely used in civil and marine engineering, pitting damage on the surface of the steel pipes brings great uncertainty to the residual strength of the structure, and there is currently no good predictive assessment method. For these steel pipe structures in service, only visual inspection or non-destructive testing methods can be used for rough assessment to formulate further remedial measures, including local reinforcement or cutting and replacement. Due to the lack of a more accurate assessment of the remaining mechanical properties, further remedial measures are often exaggerated, which invisibly increases maintenance costs. If remedial measures are not taken in a timely manner, it often leads to premature failure of the structure, posing a major safety hazard. In response to the above problems and shortcomings, it is necessary to take measures to accurately predict and analyze the mechanical properties of steel pipe structures with pitting damage during service to protect the structures throughout their service life. Summary of the Invention
[0004] The purpose of this invention is to provide a method for predicting the residual crushing mechanical properties of pitting steel pipes based on the combination of digital twins and deep learning, which can provide accurate prediction and analysis for the axial crushing mechanical performance evaluation of steel pipe structures in harsh environments.
[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:
[0006] A method for predicting the residual crush mechanical properties of pitting steel pipes based on a combination of digital twins and deep learning, the method comprising the following steps:
[0007] S1, a numerical model of steel pipes with pitting damage was established based on numerical simulation methods. The axial crushing strength of a series of steel pipes with different random pitting pit characteristics was calculated to generate a basic data set.
[0008] S2. Use a deep learning model to train and expand the basic data set, and train a large data set containing the axial crushing strength of steel pipes with different pit characteristics.
[0009] S3. For the steel pipe with pitting damage to be predicted, map it into a corresponding virtual model based on the digital twin technology, and extract the random pitting pit characteristics it contains.
[0010] S4. Based on the extracted random pitting pit characteristics, match the relevant axial crushing strength from the large data set.
[0011] Furthermore, the random pitting pit characteristics include pit depth, pit width, pit quantity, and pit distribution characteristics.
[0012] Step S1 further includes:
[0013] Collect the steel pipe structure types and pitting damage data used in civil engineering and ocean engineering.
[0014] Combined with the collected steel pipe structure types and pitting damage data, based on the finite element numerical simulation method, use the finite element software ABAQUS to establish a steel pipe structure with pit defects, set the random pit damage characteristics as variables, and establish a numerical model of a steel pipe with pitting damage.
[0015] Divide the grid for the steel pipe numerical model and set the boundary conditions, apply loads and calculate to solve its axial crushing strength.
[0016] Take different random pitting pit characteristics as independent variables and the corresponding axial crushing strength as the dependent variable to form a data set of pit characteristic parameters - axial crushing strength as the basic data set.
[0017] Step S2 further includes:
[0018] Analyze the pitting damage range on the surface of the steel pipe under the actual engineering background. Based on the basic data set in step S1, use the deep learning model to train and generate an enhanced data set containing pitting pit damage characteristics and their corresponding steel pipe axial crushing strength, and verify it on the basis of the basic data set. Take the enhanced data set as the large data set.
[0019] Step S3 further includes:
[0020] According to the structural characteristics of the steel pipe with pitting damage to be predicted, construct a physical model of the steel pipe with pitting damage, map the physical model of the steel pipe with pitting damage into a corresponding virtual mathematical model, monitor the pitting damage situation on the surface of the steel pipe in real time based on the digital twin system, and parameterize and output the pitting feature distribution on the surface of the steel pipe.
[0021] Step S4 further includes:
[0022] Based on the pitting damage characteristic distribution of the steel pipe after it has been in service for a certain period in step S3, and comparing it with the large data set trained by the deep learning model in step S2, the relationship between the corresponding pitting characteristics and the axial crushing strength is matched to predict the axial crushing strength of the steel pipe with different pitting pit damages.
[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0024] The method for predicting the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning of the present invention can be directly applied to the prediction of the mechanical properties of steel structures in the fields of civil and ocean engineering, accurately predict and analyze the mechanical properties of steel pipe structures with pitting damage during the service period, and escort the structure during its service life. Description of the Drawings
[0025] Figure 1 Schematic diagram of the numerical analysis model for the steel pipe with pitting pits;
[0026] Figure 2 Flow chart for establishing the digital twin model of the steel pipe with pitting pits;
[0027] Figure 3 Flow chart of the method for predicting the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning of the present invention. Detailed Embodiment
[0028] The following further describes the embodiments of the present invention in detail with reference to the drawings.
[0029] The present invention discloses a method for predicting the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning, and the method includes the following steps:
[0030] S1, establishing a numerical model of a steel pipe with pitting damage based on a numerical simulation method, calculating the axial crushing strength of a series of steel pipes with different random pitting pit characteristics, and generating a basic data set;
[0031] S2, using a deep learning model to train and expand the basic data set, and training a large data set containing the axial crushing strength of steel pipes with different pit characteristics;
[0032] S3, for the pitted steel pipe to be predicted, mapping it into a corresponding virtual model based on digital twin technology, and extracting the random pitting pit characteristics it contains;
[0033] S4, based on the extracted random pitting pit characteristics, matching the relevant axial crushing strength from the large data set.
[0034] See Figures 1 to 3, the prediction method of the present invention specifically includes the following steps:
[0035] S1: Considering the common steel pipe structure types in the fields of civil engineering and ocean engineering and the possible pitting corrosion damage situations, based on the finite element numerical simulation technology, select the appropriate finite element software ABAQUS, establish a steel pipe structure with a pit defect, set the random pit damage characteristics as variables, and establish a parametric numerical model of a steel pipe with random pits (as Figure 1 shown), then divide the mesh, set the boundary conditions, apply the load to calculate and solve its ultimate compressive strength. Finally, take different pit characteristic parameters as independent variables and the corresponding ultimate compressive strength as the dependent variable to form a set of basic data sets of pit characteristic parameters - compressive strength.
[0036] S2: Considering the pitting corrosion damage range on the surface of the steel pipe under the actual engineering background, based on the basic data set in step S1, use the deep learning model of the neural network algorithm to train and generate a set of enhanced data sets containing a large number of pitting corrosion pit damage characteristics and their corresponding remaining crushing strengths of the steel pipe, and verify on the basis of the basic data set, so as to evaluate the accuracy and feasibility of the deep learning model in predicting the remaining axial crushing strength of the pitted steel pipe (as Figure 2 shown).
[0037] S3: From the perspective of actual engineering, according to the characteristics of the steel pipe structure in actual service, establish a real-time monitoring digital twin system, so as to extract the pitting corrosion damage situation on the surface of the steel pipe, map the physical model of the pitted steel pipe to the corresponding virtual mathematical model, and finally based on the digital twin system, real-time monitor the pitting corrosion damage situation on the surface of the steel pipe, and parametrically output the pitting corrosion characteristic distribution on the surface of the steel pipe (as Figure 3 shown).
[0038] S4: Based on the pitting corrosion damage characteristic distribution of a specific steel pipe after serving for a certain period of time in step S3, compare with the large number of enhanced data sets trained by the deep learning model in step S2, match the corresponding pitting corrosion characteristics and compressive strength relationship, so as to accurately predict the axial crushing strength of the steel pipe with different pitting corrosion pit damages. Drawing on the operation steps of the present invention, this method is not limited to steel pipe structures, but can also be extended to the mechanical property prediction of other types of steel structures after pitting corrosion damage.
[0039] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the present application.
[0040] Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalent technologies, this application is also intended to cover these changes and modifications.
Claims
1. A prediction method for the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning, characterized in that, The method includes the following steps: S1. Based on the numerical simulation method, establish a numerical model of a steel pipe with pitting corrosion damage, calculate the axial crushing strength of a series of steel pipes with different random pitting pit characteristics, and generate a basic data set; S2. Use a deep learning model to train and expand the basic data set, and train a large data set containing the axial crushing strength of steel pipes with different pit characteristics; S3. For the steel pipe with pitting corrosion damage to be predicted, map it into a corresponding virtual model based on the digital twin technology, and extract the random pitting pit characteristics it contains; S4. Based on the extracted random pitting pit characteristics, match the relevant axial crushing strength from the large data set.
2. The method for predicting the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning according to claim 1, characterized in that, The random pitting pit characteristics include pit depth, pit width, pit quantity, and pit distribution characteristics.
3. The method for predicting the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning according to claim 1, characterized in that, Step S1 further includes: Collect the steel pipe structure types and pitting corrosion damage data used in civil engineering and ocean engineering; Combined with the collected steel pipe structure types and pitting corrosion damage data, based on the finite element numerical simulation method, use the finite element software ABAQUS to establish a steel pipe structure with pit defects, set the random pit damage characteristics as variables, and establish a numerical model of a steel pipe with pitting corrosion damage; Divide the grid for the steel pipe numerical model and set the boundary conditions, apply loads and calculate to solve its axial crushing strength; Take different random pitting pit characteristics as independent variables and the corresponding axial crushing strength as the dependent variable to form a data set of pit characteristic parameters - axial crushing strength as the basic data set.
4. The pitting steel pipe remaining crushing mechanical property prediction method based on the combination of digital twin and deep learning according to claim 1, characterized in that Step S2 further includes: Analyze the pitting corrosion damage range on the surface of the steel pipe under the actual engineering background. Based on the basic data set in step S1, with the analyzed pitting corrosion damage range as the limit, use a deep learning model to train and generate an enhanced data set containing pitting pit damage characteristics and their corresponding steel pipe axial crushing strength, and verify it on the basis of the basic data set, and use the enhanced data set as the large data set.
5. The method for predicting the remaining crushing mechanical properties of pitted steel pipes based on the combination of digital twin and deep learning according to claim 1, wherein Step S3 further includes: According to the structural characteristics of the steel pipe with pitting corrosion damage to be predicted, construct a physical model of the steel pipe with pitting corrosion damage, map the physical model of the steel pipe with pitting corrosion damage into a corresponding virtual mathematical model, monitor the pitting corrosion damage situation on the surface of the steel pipe in real time based on the digital twin system, and parameterize and output the pitting corrosion feature distribution on the surface of the steel pipe.
6. The method for predicting the remaining crushing mechanical properties of a pitted steel pipe based on the combination of digital twin and deep learning according to claim 1, wherein, [[ID=