A dynamic non-destructive prediction method and device for the residual life of an industrial equipment component

Abstract

The application discloses a kind of industrial equipment parts remaining life dynamic nondestructive prediction method and device, the application industrial equipment parts remaining life dynamic nondestructive prediction method includes: the surface roughness Ra of fatigue damage position on target part is measured;Surface roughness Ra is compared with the surface roughness Ra-residual life curve of fatigue damage position determined based on numerical simulation, to obtain the residual life of target part, the surface roughness Ra-fatigue damage position residual life curve of the mapping relationship between the surface roughness Ra of different residual life, fatigue damage position of target part is contained.The application aims to the residual life of industrial equipment parts in service process is carried out real-time, accurate nondestructive prediction, and provides data basis for the service of industrial equipment parts.

Description

Technical Field

[0001] This invention relates to the technology for detecting the remaining life of industrial equipment components, and more specifically to a dynamic non-destructive prediction method and device for the remaining life of industrial equipment components. Background Technology

[0002] With the advent of the era of intelligent manufacturing, information technology is playing an increasingly important role in the industrial field. In particular, the rise of artificial intelligence has provided effective technical means for the intelligent management of machinery and equipment in the industrial sector. During the production and service of industrial equipment, the failure of one piece of equipment or component can interrupt the production process, causing the production chain to collapse, accelerating component failure and breakage, reducing its service life, and causing significant losses. In some cases, major failures can even threaten human safety and cause irreversible consequences. Accurate prediction of the remaining service life of industrial equipment during its service is a prerequisite and necessary condition for formulating good predictive maintenance strategies. In order to predict the remaining service life of industrial equipment components during their service, reduce unnecessary accidents, and reduce economic losses and casualties, virtual simulation technology for industrial equipment component service based on digitalization, networking, and intelligence has emerged and become a research hotspot in the field of component remaining service life prediction.

[0003] Predicting the remaining service life of components has broad application prospects and market value in modern engineering, and the methods for predicting remaining service life have evolved from destructive to non-destructive approaches. However, existing methods for predicting the remaining service life of components mainly use offline methods for individual components. This approach does not consider the actual service conditions of industrial equipment components, cannot effectively predict the specific location of component failure, and lacks real-time dynamic interaction, making it difficult to perform real-time dynamic prediction of the remaining service life of industrial equipment components during their service life. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a dynamic non-destructive prediction method and device for the remaining service life of industrial equipment parts, in view of the above-mentioned problems of the prior art. The present invention aims to perform real-time and accurate non-destructive prediction of the remaining service life of industrial equipment parts, and to provide a data basis for the service of industrial equipment parts.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A dynamic non-destructive prediction method for the remaining life of industrial equipment components includes: S101, Measure the surface roughness Ra at the fatigue damage location on the target component; S102, compare the surface roughness Ra with the surface roughness Ra-remaining life curve of the fatigue damage location determined by numerical simulation to obtain the remaining life of the target component. The surface roughness Ra-remaining life curve of the fatigue damage location contains the mapping relationship between different remaining lives of the target component and the surface roughness Ra of the fatigue damage location. The generation of the surface roughness Ra-remaining life curve at the fatigue damage location includes: S201, collect data on various test parameters of component samples under standard tensile and compressive fatigue tests, including: (1) The SN curve equation of the material of the component sample is obtained by performing a standard tensile-compressive fatigue test on the material of the component sample, and the surface roughness Ra1 and damage value of the material of the component sample under different cycles are detected. The SN curve equation is the mapping relationship between the strain amplitude and the cycle of the material of the component sample. The total cycle of the component sample and the strain amplitude under different cycles are determined according to the SN curve equation. The remaining life under different cycles is obtained by subtracting the cycle from the total cycle. Finally, the surface roughness, strain amplitude, remaining life and damage value of the material of the component sample under different cycles under the standard tensile-compressive fatigue test are obtained. (2) Based on the actual service load of the target component, the typical working conditions of the component and the fatigue load under each typical working condition are extracted by numerical simulation method. The fatigue parameters obtained by standard symmetrical tensile and compressive fatigue test are used to perform numerical simulation of the service process of the component sample, determine the total number of cycles and fatigue damage location of the component sample, the surface roughness at the initial and failure of the fatigue damage location, and the peak stress, remaining life and damage value of the fatigue damage location at different number of cycles. S202, using the surface roughness of the fatigue damage location under different cycles as the output parameter and the remaining experimental parameters as the input parameter, constructs a training dataset and uses the training dataset to train a machine learning model; S203, using a trained machine learning model to predict the surface roughness Ra under different remaining lifespans, thereby obtaining the surface roughness Ra-remaining life curve at the fatigue damage location of the target component.

[0006] Optionally, measuring the surface roughness Ra at the fatigue damage location on the target component in step S101 means measuring the surface roughness Ra at the fatigue damage location on the target component using a white light interferometric roughness measuring instrument.

[0007] Optionally, in step S202, when training the machine learning model using the training dataset, the process includes constructing multiple different machine learning models, using multiple different loss functions for each machine learning model, training each combination of machine learning model and loss function using the training dataset, testing the accuracy of each combination of machine learning model and loss function, and finally selecting the combination of machine learning model and loss function with the best accuracy as the trained machine learning model.

[0008] Optionally, the various machine learning models include some or all of the following: support vector machine model, K-nearest neighbor model, Naive Bayes model, gradient boosting model, and random forest model.

[0009] Optionally, the various loss functions include one of mean squared error, root mean square error, and mean absolute error.

[0010] Furthermore, the present invention also provides a dynamic non-destructive prediction device for the remaining life of industrial equipment components, comprising a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the dynamic non-destructive prediction method for the remaining life of the industrial equipment components.

[0011] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program for being programmed or configured by a microprocessor to perform a dynamic non-destructive prediction method for the remaining life of industrial equipment components.

[0012] Compared with the prior art, the present invention has the following main advantages: 1. The dynamic non-destructive prediction method for the remaining service life of industrial equipment parts of the present invention can realize real-time and accurate prediction of the remaining service life of industrial equipment parts during service, providing a data basis for the service of industrial equipment parts.

[0013] 2. This invention can achieve real-time non-destructive testing based on the surface roughness Ra at the main damage location (fatigue damage location). By comparing the surface roughness Ra-remaining life curve at the fatigue damage location, dynamic non-destructive prediction of the remaining life of the component can be achieved. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the basic process of the method in an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of the process for generating the surface roughness Ra-remaining lifetime curve in an embodiment of the present invention.

[0016] Figure 3 The true stress-true plastic strain curves are for multiple grades of aluminum alloys in the embodiments of the present invention.

[0017] Figure 4 This is the strain amplitude-fatigue failure life curve in an embodiment of the present invention. Detailed Implementation

[0018] The following description uses the front bumper beam assembly with an aluminum alloy extruded profile structure as an example to further illustrate the invention. The total weight of the front bumper beam assembly is 3.2 kg. The front beam and tow hook sleeve are made of 6061-T6 aluminum alloy, while the energy-absorbing box, base plate, and bracket are made of 6082-T6 aluminum alloy.

[0019] like Figure 1 As shown, the dynamic non-destructive prediction method for the remaining life of industrial equipment components in this embodiment includes: S101, Measure the surface roughness Ra at the fatigue damage location on the target component; S102, the surface roughness Ra is compared with the surface roughness Ra-remaining life curve of the fatigue damage location determined by numerical simulation, so as to obtain the remaining life of the target component. The surface roughness Ra-remaining life curve of the fatigue damage location contains the mapping relationship between different remaining lives of the target component and the surface roughness Ra of the fatigue damage location.

[0020] In step S101 of this embodiment, measuring the surface roughness Ra at the fatigue damage location on the target component refers to measuring the surface roughness Ra at the fatigue damage location on the target component using a white light interferometric roughness measuring instrument.

[0021] like Figure 2 As shown, before step S102, the method further includes generating the surface roughness Ra-remaining life curve at the fatigue damage location: S201, Collect data on various test parameters of the component samples under standard tensile and compressive fatigue tests and construct a training dataset. The test parameters include the surface roughness, strain amplitude, remaining life and damage value of the material used in the component samples under different cycles of standard tensile and compressive fatigue tests, the fatigue damage location of the component samples, the strain amplitude, remaining life and damage value of the fatigue damage location under different cycles, and the surface roughness data of the component samples at the initial and failure times. S202, using the surface roughness of the fatigue damage location under different cycles as the output parameter and the remaining experimental parameters as the input parameter, constructs a training dataset and uses the training dataset to train a machine learning model; S203, using a trained machine learning model to predict the surface roughness Ra under different remaining lifespans, thereby obtaining the surface roughness Ra-remaining life curve at the fatigue damage location of the target component.

[0022] In this embodiment, step S201, which involves collecting data on various test parameters of the component sample under standard tensile-compressive fatigue testing, includes: conducting standard tensile-compressive fatigue testing on the material used in the component sample to obtain the SN curve equation of the material of the component sample, and detecting the surface roughness Ra1 and damage value of the material of the component sample under different cycles. The SN curve equation is a mapping relationship between the strain amplitude and the number of cycles of the material of the component sample. Based on the SN curve equation, the total number of cycles of the component sample and the strain amplitude under different cycles are determined, and the remaining life under different cycles is obtained by subtracting the number of cycles from the total number of cycles. Finally, the surface roughness, strain amplitude, remaining life and damage value of the material used in the component sample under different cycles under standard tensile-compressive fatigue testing are obtained, as shown in Table 1.

[0023] Table 1: Data from standard tensile and compressive fatigue tests.

[0024] In the table above, N represents the total number of cycles, Δσ represents the stress, and Δε represents the stress. t The value is the strain amplitude, and the step size of the cycle is 2 (which can be selected according to actual needs).

[0025] In this embodiment, step S201, when collecting data on various test parameters of the component sample under the standard tensile and compressive fatigue test, includes: based on the actual service load of the target component, extracting the typical working conditions of the component and the fatigue load under each typical working condition using numerical simulation methods, and using the fatigue parameters obtained through the standard symmetrical tensile and compressive fatigue test to perform numerical simulation of the service process of the component sample, determining the total number of cycles and fatigue damage location of the component sample, the surface roughness at the initial fatigue damage location and at failure, and the peak stress, remaining life and damage value at different cycle numbers of the fatigue damage location, as shown in Table 2.

[0026] Table 2: Numerical simulation data for any fatigue damage location k.

[0027] In the above table, n k σ represents the total number of cycles at fatigue damage location k. k ε is the peak stress at fatigue damage location k. k Let K be the strain at fatigue damage location k, and let the cycle size of fatigue damage location k be 2 (which can be selected according to actual needs).

[0028] In this embodiment, the tensile strain rate during the standard tensile-compression fatigue test is 0.01 s⁻¹. -1Three qualified test data points were taken at each strain rate and normalized. The true stress-true plastic strain curves for three grades of aluminum alloys, 6061-T6, 6082-T6, and 6063-T6, are shown below. Figure 3 As shown. Based on quasi-static mechanical parameters, numerical simulation was used to obtain the stress distribution of the components, determine the high-stress regions under various working conditions, and thus determine the fatigue damage location. Then, strain control was adopted using standard tensile-compressive fatigue, with strain amplitudes of 0.002, 0.004, 0.006, 0.008, and 0.01, and strain rates of 6 × 10⁻⁶. -3 s -1 Three qualified test data points were taken for each strain amplitude. The strain amplitude-fatigue failure life curve of a certain material is shown below. Figure 4 As shown, the figure includes the total strain amplitude, elastic strain amplitude, and plastic strain amplitude, respectively. Based on the standard tensile and compressive fatigue parameters of the material, numerical simulations of fatigue damage under combined fatigue conditions can determine the fatigue damage values ​​and fatigue life of the main fatigue damage locations of the components. Using a white light interferometric roughness measuring instrument, the initial (cycle count 0) and post-fatigue test (cycle count n) values ​​of high-stress areas and fatigue damage locations can be acquired. k ) Surface roughness Ra database.

[0029] Fatigue life N f The fatigue life of a material or component refers to the number of cyclic deformations it undergoes before fatigue failure; 2N f The number of cyclic reversals that have occurred by the time of cyclic failure is represented, and its value is twice the fatigue life. Remaining fatigue life refers to the fatigue life of a material or component after it has been operating under certain cyclic loading conditions for a cumulative period of time, continuing to operate under those conditions until failure. Its value is the difference between the fatigue life of the material or component under those conditions and its cumulative service life. Total strain amplitude Δε t / 2 refers to half the difference between the maximum and minimum total strain during cyclic deformation, where tensile strain is positive and compressive strain is negative. The total strain amplitude is equal to the sum of the plastic strain amplitude and the elastic strain amplitude, denoted by the symbol ε. t ε p and ε e These represent the total strain, plastic strain, and elastic strain, respectively. Elastic strain amplitude Δε e / 2 refers to half the difference between the maximum and minimum elastic strain during cyclic deformation; tensile strain is positive, and compressive strain is negative. Plastic strain amplitude Δε p / 2 refers to half the difference between the maximum and minimum plastic strain during cyclic deformation. Tensile strain is positive, and compressive strain is negative. In the fatigue life curve, the horizontal axis (x) represents the fatigue life under constant strain amplitude loading. It is a commonly used logarithmic function with the power (property) as the independent variable, the exponent as the dependent variable, and a base of 10. Its value is: x = lg2Nf The vertical axis (y) represents the strain amplitude Δε (total strain amplitude Δε). t / 2, Plastic strain amplitude Δε p / 2 and elastic strain amplitude Δε e / 2), also a common logarithmic function with the power (property) as the independent variable, the exponent as the dependent variable, and a base of 10, has the following values: y=lgΔε t / 2, y=lgΔε p / 2 and y=lgΔε e / 2.

[0030] In this embodiment, step S202, when training the machine learning model using the training dataset, includes constructing multiple different machine learning models and employing multiple different loss functions for each model. The training dataset is used to train each combination of machine learning model and loss function, and the accuracy of each combination is tested. Finally, the combination of machine learning model and loss function with the best accuracy is selected as the trained machine learning model. In this embodiment, the training dataset is imported into Python, and the data is randomly divided into training data (80%) and test data (20%) through programming. The Z-Score standardization method is used to transform and map the original data samples into data with a mean of 0 and a standard deviation of 1. Then, the training dataset is used to train each combination of machine learning model and loss function, and the accuracy of each combination is tested. Finally, the combination of machine learning model and loss function with the best accuracy is selected as the trained machine learning model. Input test data is used to predict property indicators. Hyperparameters such as learning rate, number of iterations, and batch size are optimized during model training to improve the prediction accuracy and generalization ability of the machine learning model. This allows the constructed machine learning model to generate surface roughness Ra-remaining life curves for fatigue damage locations. The various machine learning models include, but are not limited to, some or all of the following: Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Naive Bayes, Gradient Boosting (GPRB), and Random Forest. The various loss functions include, but are not limited to, one of the following: Mean Squared Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE).

[0031] In summary, the dynamic non-destructive prediction method for the remaining service life of industrial equipment components in this embodiment can accurately predict the remaining service life of industrial equipment components by performing real-time non-destructive testing on the surface roughness Ra parameter of key parts of industrial equipment components during service, and comparing the surface roughness Ra-remaining service life curve of standardized fatigue damage locations.

[0032] Furthermore, this embodiment also provides a dynamic non-destructive prediction device for the remaining life of industrial equipment components, including a microprocessor and a memory interconnected thereto. The microprocessor is programmed or configured to execute the aforementioned dynamic non-destructive prediction method for the remaining life of industrial equipment components. Additionally, this embodiment also provides a computer-readable storage medium storing a computer program for being programmed or configured by the microprocessor to execute the aforementioned dynamic non-destructive prediction method for the remaining life of industrial equipment components.

[0033] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer-readable storage media. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer programs according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create an implementation for the process. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0034] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A dynamic non-destructive prediction method for the remaining life of industrial equipment components, characterized in that, include: S101, Measure the surface roughness Ra at the fatigue damage location on the target component; S102, compare the surface roughness Ra with the surface roughness Ra-remaining life curve of the fatigue damage location determined by numerical simulation to obtain the remaining life of the target component. The surface roughness Ra-remaining life curve of the fatigue damage location contains the mapping relationship between different remaining lives of the target component and the surface roughness Ra of the fatigue damage location. The generation of the surface roughness Ra-remaining life curve at the fatigue damage location includes: S201, collect data on various test parameters of component samples under standard tensile and compressive fatigue tests, including: (1) The SN curve equation of the material used in the component sample is obtained by performing a standard tensile-compressive fatigue test on the material. The surface roughness Ra1 and damage value of the material used in the component sample are detected at different cycles. The SN curve equation is the mapping relationship between the strain amplitude and the number of cycles of the material in the component sample. The total number of cycles and the strain amplitude at different cycles are determined according to the SN curve equation. The remaining life at different cycles is obtained by subtracting the number of cycles from the total number of cycles. Finally, the surface roughness Ra1, strain amplitude, remaining life and damage value of the material used in the component sample at different cycles under the standard tensile-compressive fatigue test are obtained. The strain amplitude is equal to the plastic strain amplitude Δε. p / 2 and elastic strain amplitude Δε e The sum of / 2, elastic strain amplitude Δε e / 2 refers to half the difference between the maximum and minimum elastic strain during cyclic deformation; tensile strain is positive, and compressive strain is negative; the plastic strain amplitude Δε p / 2 refers to half the difference between the maximum and minimum plastic strain during cyclic deformation, with tensile strain being positive and compressive strain being negative; the remaining life refers to the fatigue life of a material or component after it has been running under certain cyclic load conditions for a period of time, continuing to operate under those conditions until failure, and its value is the fatigue life N of the material or component under those conditions. f The fatigue life N is the difference between the cumulative service life and the cumulative service life. f The number of cycles of deformation that a material or component undergoes when it fails due to fatigue. (2) Based on the actual service load of the target component, the typical working conditions of the component and the fatigue load under each typical working condition are extracted by numerical simulation method. The fatigue parameters obtained by standard symmetrical tensile and compressive fatigue test are used to perform numerical simulation of the service process of the component sample, determine the total number of cycles and fatigue damage location of the component sample, the surface roughness Ra1 at the initial and failure of the fatigue damage location, and the peak stress, remaining life and damage value of the fatigue damage location at different number of cycles. S202, using the surface roughness of the fatigue damage location under different cycles as the output parameter and the remaining experimental parameters as the input parameters, constructs a training dataset and uses the training dataset to train a machine learning model, including: constructing multiple different machine learning models, using multiple different loss functions for each machine learning model, training each machine learning model and loss function combination using the training dataset, and finally testing the accuracy of each machine learning model and loss function combination, and finally selecting the machine learning model and loss function combination with the best accuracy as the trained machine learning model; S203, using a trained machine learning model to predict the surface roughness Ra under different remaining lifespans, thereby obtaining the surface roughness Ra-remaining life curve at the fatigue damage location of the target component.

2. The dynamic non-destructive prediction method for the remaining life of industrial equipment components according to claim 1, characterized in that, In step S101, measuring the surface roughness Ra at the fatigue damage location on the target component refers to measuring the surface roughness Ra at the fatigue damage location on the target component using a white light interferometric roughness measuring instrument.

3. The dynamic non-destructive prediction method for the remaining life of industrial equipment components according to claim 1, characterized in that, In step S202, when training a machine learning model using a training dataset, the process includes constructing multiple different machine learning models, using multiple different loss functions for each machine learning model, training each combination of machine learning model and loss function using the training dataset, testing the accuracy of each combination of machine learning model and loss function, and finally selecting the combination of machine learning model and loss function with the best accuracy as the trained machine learning model.

4. The dynamic non-destructive prediction method for the remaining life of industrial equipment components according to claim 3, characterized in that, The various machine learning models include some or all of the following: support vector machine model, K-nearest neighbor model, Naive Bayes model, gradient boosting model, and random forest model.

5. The dynamic non-destructive prediction method for the remaining life of industrial equipment components according to claim 3, characterized in that, The various loss functions include one of mean squared error, root mean square error, and mean absolute error.

6. A dynamic non-destructive prediction device for the remaining life of industrial equipment parts, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to execute the dynamic non-destructive prediction method for the remaining life of industrial equipment components as described in any one of claims 1 to 5.

7. A computer-readable storage medium storing a computer program, characterized in that, The computer program is used to be programmed or configured by a microprocessor to execute the dynamic non-destructive prediction method for the remaining life of industrial equipment components as described in any one of claims 1 to 5.

Images