CN107391890B - Prediction and optimal control method for oil casing threaded joint machining chatter defect - Google Patents

Abstract

The invention discloses a prediction and optimization control method for the machining chatter pattern defect of an oil casing threaded joint, and relates to the technical field of petroleum equipment manufacturing. The invention aims at the law and characteristics of chatter pattern occurrence in thread machining and is based on the actual chatter pattern occurrence situation. Big data, cross-application of data mining, reliability analysis and optimal design theory methods, especially using the improved K-means clustering algorithm to implement data mining, using the improved non-dominated sorting genetic algorithm to implement the optimization of chatter pattern defect control, and proposed the oil casing Prediction and optimization control method for machining chatter pattern defects of threaded joints; compared with the prior art, the invention provides a prediction and optimization control method for machining chatter pattern defects of oil casing threaded joints, which can improve the machining of oil casing threads. The accuracy of chatter pattern prediction and the effective control of chatter pattern occurrence.

Description

A prediction and optimization control method for machining chatter pattern defects of oil casing threaded joints

technical field

The invention relates to the technical field of petroleum equipment manufacturing, in particular to a prediction and optimization control method for the machining chatter pattern defect of an oil casing threaded joint.

Background technique

The oil and gas well tubing string is made up of a large number of tubing and casing threads connected by a single thread, and the number of tubing and casing threads in a single oil and gas well is often in the thousands. Thread is an important link in the service structure and sealing integrity assurance of the pipe string under composite loads such as downhole tension, compression, bending, torsion, internal/external pressure, etc., and it is also the weakest link (according to incomplete statistics, downhole pipe string structure and About 70% of seal integrity failures occur in the threaded portion). The harsh service environment of the thread and the extremely high safety and reliability index put forward higher requirements on the anti-galling, connection performance and sealing performance of the thread. The machining accuracy and quality of the thread are very important.

The chatter pattern is a common phenomenon in the production and processing of special threaded joints. It not only affects the surface quality of special threaded joints, but also leads to the deterioration of various properties of special threaded joints, directly reducing the performance of the thread. The effect of sealing performance is particularly significant. The occurrence of chatter lines is affected by the complex coupling of many factors, including: tube body overhang length, inner support and floating support rollers related to tube stiffness; tool shank diameter, overhang length related to tool stiffness, and lathe rigidity Relevant bearing clearance, bearing lubricity data; clamping pressure, jaw size, clamping contact area related to the way of sample clamping; blade and shim clearance related to tool quality and installation, clamping tightness, The high distance between the tool nose and the center of the lathe; the depth of cut, the amount of passes, etc. related to the machining process. In the prior art, repeated adjustment and optimization are often carried out through process trial and error, which leads to the difficulty in finding the cause of chatter patterns and the lack of avoidance measures in engineering practice. Aiming at the law and characteristics of the occurrence of chatter lines in thread processing, based on the actual big data of chatter lines occurrence, the invention crosses the theoretical methods of data mining, reliability analysis and optimization design, and proposes prediction of the machining chatter pattern defects of oil casing threaded joints and optimized control methods.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a prediction and optimization control method for the machining chatter pattern defect of an oil casing threaded joint, so as to improve the accuracy of chatter pattern occurrence prediction and effectively control chatter pattern generation.

The technical scheme adopted by the present invention to solve the technical problem is: a prediction and optimization control method for machining chatter pattern defects of oil casing threaded joints, which is characterized in that a plurality of groups of different steel grades, specifications and The thread machining process parameters of the thread type and the corresponding actual chatter pattern data on the thread surface; the machining process parameter data includes the overhang length of the pipe body, the inner support and the floating support roller related to the stiffness of the pipe sample, and the tool stiffness related to the tool stiffness. Rod diameter, overhang length, bearing clearance related to lathe rigidity, bearing lubricity data, clamping pressure, jaw size, clamping contact area related to sample clamping method, blade and Tool shim clearance, clamping tightness, high distance between the tool tip and the center of the lathe; the depth of cut, the amount of tool passes, etc. related to the machining process; Rating data for degree observation and detection; the method includes:

S1, select at least two groups of oil casing threading process parameter data of the same specification, steel grade and thread type stored in the original database and their corresponding actual thread surface chatter pattern data, and normalize and abstract the selected data is a vector, divide the normalized data into training data samples and verification data samples, and put the training data samples and verification data samples into the training sample database and the verification sample database respectively;

S2, using the improved K-means clustering algorithm to perform data mining modeling on the training data samples, and obtain the data mining-based data mining of oil casing thread processing with the machining process parameters as the dependent variable and the thread surface chatter pattern data as the response variable The chatter pattern production prediction model;

S3, use the test data sample to verify the correctness of the chatter pattern production prediction model based on data mining, if the test is passed and the chatter pattern prediction expert model is obtained, then execute S4, otherwise, transpose S1, select the updated data to repeat Implementation process;

S4, using the obtained expert model for predicting the chattering pattern to predict the degree of chattering pattern generation under the specified process conditions input by the user, and outputting the chattering pattern prediction result;

S5, according to a large number of actual measured machining process parameters in the training sample library, analyze and obtain a process parameter distribution model, and combine the chatter pattern prediction expert model established in S4 to establish a response surface function-based chatter pattern reliability prediction model through Monte Carlo simulation;

S6, the target reliability of the design is determined through the comprehensive analysis of the analysis of the rejection rate requirement of thread processing and the analysis of the cost control requirement, and the target reliability is used as the standard as the target reliability check, and the reliability of the chatter pattern calculated by S5 is used. The difference between the flutter reliability and the target reliability in the prediction model is used as the objective function, and the processing parameters are used as the design variables. The improved non-dominated sorting genetic algorithm is used to perform reliability optimization iteration, and finally the target reliability requirements are met. Output Optimal design results of process parameters controlling chatter pattern.

In the described S2, data mining modeling is implemented according to the principle of the K-means clustering algorithm, and the described data mining modeling steps are:

(1) Arbitrarily select k objects as the initial cluster centers from the mathematical vectors of n process parameters that actually form the chatter pattern;

(2) Calculate the distance between each object and the center of these clusters according to the mean of the objects in the cluster, and assign each object to the most similar cluster;

(3) Update the cluster mean, that is, calculate the mean of the objects in each cluster;

(4) Loop (2) and (3) until each cluster criterion function no longer changes;

Completed prediction of multiple process parameters for chatter pattern formation.

As an optimization, the K-means clustering algorithm is based on the principle of selecting K points with a batch distance as far as possible as the initial cluster. First, a point is randomly selected as the center point of the first initial cluster, and then the distance from this point is selected. The farthest point is used as the center point of the second initial cluster, and then the point with the largest distance from the first two points is selected as the center point of the third initial cluster, and so on, until K objects are selected as the initial cluster center.

The reliability prediction model of the tremor pattern uses the improved non-dominated sorting genetic algorithm for optimization iteration, and the reliability optimization iteration adopts the optimal design of the non-dominated sorting genetic algorithm. The implementation includes genetic code generation, random initial population determination, fitness The optimization iterative process of evaluation and calculation of the offspring population, on the basis of evaluating the Pareto ranking value of the individual, by introducing the mean and standard deviation of the individual variable and its ranking value to evaluate the optimization target approach to the population density information around the individual. susceptibility to keep the offspring of individuals away from areas where the population is inactive.

The beneficial effects of the invention are: based on the actually detected thread machining process parameter data and the corresponding thread machining actually generated chatter pattern data, a chatter pattern prediction model based on data mining is established by the improved K-means method, and the preset chatter pattern is determined according to the model. The occurrence degree of chatter lines under the condition of process parameters is predicted; combined with the statistics of the distribution data of thread processing parameters, the reliability prediction model of chatter lines is established by Monte Carlo simulation, and the expected chatter lines reliability indicators are established according to the consideration of quality and cost. The improved non-dominated sorting genetic algorithm takes the expected chatter pattern reliability index as the optimization goal, and takes the thread machining process parameters as the optimized design parameters, and performs the optimization calculation to obtain the optimization results of the machining process parameters that control the chatter pattern occurrence. Therefore, the accuracy of the prediction of the occurrence of chatter patterns can be improved, and the occurrence probability of chatter patterns can be effectively controlled.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

Fig. 1 is a flow chart of the control method of the present invention.

Detailed ways

The principles and features of the present invention will be described below with reference to the accompanying drawings. The examples are only used to explain the present invention, but not to limit the scope of the present invention.

A prediction and optimization control method for machining chatter pattern defects of oil casing threaded joints mainly includes the following steps:

1. The user establishes a database of the actual chatter pattern of oil casing thread processing, and inputs and continuously accumulates, updates and stores multiple sets of thread machining process parameters of different steel grades, specifications and thread types and the corresponding actual chatter pattern data on the thread surface in the database; The machining process parameter data includes: tube body overhang length, internal support and floating support rollers related to the stiffness of the tube sample, tool shank diameter and overhang length related to the tool stiffness, bearing clearance and bearing lubricity related to the rigidity of the lathe Data; clamping pressure, jaw size, clamping contact area related to the way of sample clamping; blade and shim clearance related to tool quality and installation, clamping tightness, high distance between the tool tip and the center of the lathe; The depth of cut, the amount of passes, etc. related to the machining process. Actual chatter pattern data on the thread surface, including: rating data for observation and detection of the severity of chatter pattern on the actual machined thread surface.

2. Normalize and abstract at least two sets of oil casing threading process parameter data of the same specification, steel grade and thread type stored in the original database and the corresponding actual surface chatter pattern data of the thread into a vector. The quantized data is divided into training data samples and verification data samples, and the training data samples and verification data samples are put into the training sample database and the verification sample database respectively.

3. Using the improved K-means clustering algorithm, the training data samples are modeled by data mining, and the data mining based on the thread machining of the oil casing with the machining process parameters as the dependent variable and the thread surface chatter pattern data as the response variable is obtained. A prediction model for chatter pattern production.

Further, the chatter pattern is affected by the complex coupling of the above-mentioned multi-dimensional data, and it is difficult to determine the internal mechanism. Implement data mining modeling according to the principle of K-means clustering algorithm:

(1) Arbitrarily select k objects as the initial cluster centers from the mathematical vectors of n process parameters that actually form the chatter pattern;

(2) Calculate the distance between each object and the center of these clusters according to the mean of the objects in the cluster, and assign each object to the most similar cluster;

(3) Update the cluster mean, that is, calculate the mean of the objects in each cluster;

(4) Loop (2) and (3) until each cluster criterion function no longer changes;

Thus, the prediction of multiple process parameters for forming chatter patterns is completed;

Further, the improvement of the K-means clustering algorithm: The K-means clustering algorithm is particularly sensitive to the initial value of the initial cluster center and the number of selections, and an improvement is proposed for this problem: based on the selection of the K clusters that are as far away from the batch as possible. The principle of point as the initial cluster, first randomly select a point as the center point of the first initial cluster, then select the point farthest from the point as the center point of the second initial cluster, and then select the distance from the first two points The point with the largest closest distance is used as the center point of the third initial cluster, and so on, until K objects are selected as the initial cluster center.

4. Use the test data sample to verify the correctness of the data mining-based chatter pattern production prediction model. If the test is passed, the chatter pattern prediction expert model is obtained, then the chatter pattern prediction is performed, and the chatter pattern generated under the specified process conditions is generated. Predict the degree and output the prediction result, otherwise, transpose the original database of the actual chatter pattern of oil casing thread processing, and select the updated database to repeat the process.

5. According to a large number of measured machining process parameters in the training sample library, analyze and obtain the process parameter distribution model, and combine the established chatter pattern prediction expert model to establish a response surface function-based chatter pattern reliability prediction model through Monte Carlo simulation.

6. Determine the target reliability of the design through the comprehensive analysis of the thread processing rejection rate requirement analysis and the cost control requirement analysis.

7. Use the established response surface function-based chatter pattern reliability prediction model to calculate the chatter pattern reliability, compare the difference with the target reliability as the objective function, use the machining process parameters as the design variables, and use the improved non-dominated sorting genetic The algorithm performs reliability optimization iteration, finally meets the requirements of target reliability, and outputs the optimization design results of process parameters that control the degree of chattering.

Further, according to the optimization design idea of the non-dominated sorting genetic algorithm, an optimization iterative process including genetic code generation, random initial population determination, fitness evaluation and calculation of the offspring population is implemented. Among them, although the fitness evaluation process can obtain and evaluate the Pareto ranking value of all genetic individuals through non-dominated sorting, so as to continuously approach the optimal solution, it cannot evaluate the population density information around individuals at different levels, so sometimes there may be different optimization goals. The individuals still have the same chance to reproduce offspring, thus reducing the optimization efficiency. Aiming at this deficiency of the original algorithm, an improvement is proposed: on the basis of evaluating the Pareto ranking value of the individual, by introducing the mean and standard deviation of the individual variable and its ranking value to evaluate the effect of the optimization target approximation on the population density information around the individual. Sensitivity, so that the next generation of individuals is far away from the inactive area of the population, which increases the rationality of fitness evaluation and improves the efficiency of optimization iteration.

Below in conjunction with embodiment, the inventive method is described in detail:

In the embodiment shown in FIG. 1 , a prediction and optimization control method for machining chatter pattern defects of an oil casing threaded joint is provided. The user establishes the original database of the actual chatter pattern of the thread processing of the tubing and casing, and inputs multiple sets of thread machining process parameters of different steel grades, specifications and thread types and the corresponding actual chatter pattern data on the thread surface collected in the actual production process into it. The processing parameter data includes: the overhang length of the pipe body, the inner support and the floating support roller related to the rigidity of the pipe sample, the diameter and the overhang length of the tool shank related to the rigidity of the tool, and the bearing clearance and bearing lubrication related to the rigidity of the lathe. performance data; clamping pressure, jaw size, clamping contact area related to sample clamping method; blade and shim clearance, clamping tightness, high distance between tool tip and lathe center related to tool quality and installation ; The depth of cut, the amount of passes, etc. related to the machining process. Actual chatter pattern data on the thread surface, including: rating data for observation and detection of the chatter pattern severity on the actual machined thread surface.

Normalize and abstract at least two sets of oil casing threading process parameter data of the same specification, steel grade and thread type stored in the original database and their corresponding actual surface chatter pattern data into vectors. The data is divided into training data samples and verification data samples, and the training data samples and verification data samples are put into the training sample database and the verification sample database respectively;

Using the established improved K-means clustering algorithm, the training data samples are modeled by data mining, and the test data samples are used to check the correctness of the model. Based on the actual large database, the machining process parameters are obtained as dependent variables. , Thread surface chatter pattern data as a variable for oil casing thread machining chatter pattern prediction model;

Predict the degree of chatter pattern generation under the specified process conditions, and output the prediction result;

According to a large number of measured machining process parameters in the training sample library, analyze and obtain the process parameter distribution model, and combine the established chatter pattern prediction expert model to establish the chatter pattern reliability prediction model based on the response surface function through Monte Carlo simulation;

Determine the target reliability of the design through the comprehensive analysis of the thread processing rejection rate and cost control,

The chatter texture reliability is calculated by the established response surface function-based chatter texture reliability prediction model, the difference value compared with the target reliability is used as the objective function, the processing parameters are used as design variables, and the established improved non-dominated sorting genetic algorithm is used. Carry out optimization iterations, finally meet the requirements of target reliability, and output the optimization design results of process parameters that control the degree of chatter patterns.

The above-mentioned specific embodiments are only specific cases of the present invention, and the patent protection scope of the present invention includes but is not limited to the product forms and styles of the above-mentioned specific embodiments. The method for predicting and optimizing control of grain defects and any appropriate changes or modifications made by those skilled in the art shall fall within the scope of patent protection of the present invention.

Claims (2)

1. A prediction and optimization control method for machining chatter pattern defects of oil casing threaded joints, characterized in that, in the original database, multiple groups of thread machining process parameters and corresponding threads of different steel grades, specifications and thread types are stored in advance The actual chatter pattern data on the surface; the processing parameter data includes the overhang length of the pipe body, the inner support and the floating support roller related to the rigidity of the pipe sample, the diameter of the shank and the overhang length related to the rigidity of the tool, and the rigidity of the lathe Bearing clearance, bearing lubricity data, clamping pressure, jaw size, clamping contact area related to sample clamping method, blade and shim clearance related to tool quality and installation, clamping tightness, tool The high distance between the tip and the center of the lathe; the depth of cut and the amount of tool passes related to the machining process; the actual chatter pattern data on the thread surface includes, the rating data for the observation and detection of the severity of the chatter pattern on the surface of the actually machined thread; the method includes : S1, select at least two groups of oil casing threading process parameter data of the same specification, steel grade and thread type stored in the original database and their corresponding actual thread surface chatter pattern data, and normalize and abstract the selected data is a vector, divide the normalized data into training data samples and verification data samples, and put the training data samples and verification data samples into the training sample database and the verification sample database respectively; S2, using the improved K-means clustering algorithm to perform data mining modeling on the training data samples, and obtain the data mining-based data mining of oil casing thread processing with the machining process parameters as the dependent variable and the thread surface chatter pattern data as the response variable The K-means clustering algorithm is based on the principle of selecting the K points with the farthest batch distance as the initial cluster, first randomly select a point as the center point of the first initial cluster, and then select The point that is farthest from this point is used as the center point of the second initial cluster, and then the point with the largest distance from the first two points is selected as the center point of the third initial cluster, and so on, until the selection is made. K objects are used as initial cluster centers; S3, using the verification data sample to verify the correctness of the data mining-based chatter pattern production prediction model, if the test is passed, the chatter pattern prediction expert model is obtained, then perform S4, otherwise, transpose S1, and select the updated data Repeat the process; S4, using the obtained expert model for predicting the chattering pattern to predict the degree of chattering pattern generation under the specified process conditions input by the user, and outputting the chattering pattern prediction result; S5, according to a large number of actual measured machining process parameters in the training sample library, analyze and obtain a process parameter distribution model, and combine the chatter pattern prediction expert model established in S4 to establish a response surface function-based chatter pattern reliability prediction model through Monte Carlo simulation; S6, the target reliability of the design is determined through the comprehensive analysis of the analysis of the rejection rate requirement of the thread processing and the analysis of the cost control requirement, and the target reliability is used as the standard as the target reliability check, and the reliability of the chatter pattern calculated by S5 is used. The difference between the flutter reliability and the target reliability in the prediction model is used as the objective function, and the machining process parameters are used as design variables. The improved non-dominated sorting genetic algorithm is used to perform reliability optimization iterations, and finally meet the target reliability requirements. Output Optimal design results of process parameters for controlling tremor; the specific implementation of the improved non-dominated sorting genetic algorithm includes the optimization iterative process of genetic code generation, random initial population determination, fitness evaluation, and calculation of progeny population. On the basis of evaluating the individual variables and the mean and standard deviation of their ranking values, the sensitivity of the optimization target approach to the population density information around the individual is evaluated, so that the next generation of the individual is far away from the inactive area of the population. 2. the prediction and optimization control method of a kind of oil casing threaded joint machining chatter pattern defect as claimed in claim 1, it is characterized in that, in described S2, implement data according to the principle of improved K-means clustering algorithm Mining modeling, the described data mining modeling steps are: (1) Arbitrarily select k objects as the initial cluster centers from the mathematical vectors of n process parameters that actually form the chatter pattern; (2) Calculate the distance between each object and the center of these clusters according to the mean of the objects in the cluster, and assign each object to the most similar cluster; (3) Update the cluster mean, that is, calculate the mean of the objects in each cluster; (4) Loop (2) and (3) until each cluster criterion function no longer changes; Completed prediction of multiple process parameters for chatter pattern formation.

Images