A method for predicting and analyzing big data of drilling tool life
The life prediction model of drilling tool is established through Bayesian inference algorithm, Markov process analysis method and artificial neural network algorithm, which solves the problem of life prediction of drilling tool in the existing technology, realizes accurate prediction and fault detection of drilling tool life, and reduces engineering Cost and risk improve work efficiency and material management efficiency.
Patent Information
- Application Number
- CN202111117794.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-24
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-09-24
AI Technical Summary
The existing technology is difficult to effectively predict and analyze the life of drilling tools, resulting in waste of resources and high engineering costs, and it is difficult to reduce drilling risks.
The data of drilling tool is analyzed and predicted by Bayesian inference algorithm, Markov process analysis method and artificial neural network algorithm (ANN), and a fully interconnected feedforward hidden layer network model is established to predict drilling tool life and fault detection, providing Early warning of potential failures.
It realizes accurate prediction of drilling tool life, reduces engineering costs, improves the reliability and stability of drilling systems, reduces tool failures, and improves work efficiency and material management efficiency.
Smart Images

Figure CN114462662B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil and natural gas well completion, and more specifically to a method for predicting and analyzing big data on the life of a drilling tool. Background Art
[0002] Big data has the new characteristics of scale, high speed, diversity, and ubiquity. Specifically, it means that it is necessary to quickly obtain, process, analyze and extract valuable, massive and diverse transaction data and interaction data as the basis to propose targeted solutions for the operation mode of the enterprise. Through the collection and analysis of performance data during the operation of drilling tools, the use of computer software systems to detect and analyze equipment data, and finally determine the equipment operation status model, quantitative analysis is made for equipment procurement, use, maintenance, and rotation. Through continuous data analysis, the service life of equipment is reasonably determined, production is scientifically and rationally organized, and enterprises are guided to make more scientific and effective judgments to avoid waste of resources.
[0003] The research of prediction technology is an important aspect of scientific research. It can make scientific and effective decisions based on the prediction results. The development of artificial intelligence technology has driven the research of prediction technology, and the data-driven modeling intelligent prediction method has attracted widespread attention. As a specific application of prediction technology, life prediction has become a hot topic in the research of electromechanical equipment. It builds a life model based on the data of the monitored equipment and judges the status and life of the equipment through the model results. Summary of the invention
[0004] The present invention overcomes the deficiencies in the prior art and provides a method for predicting and analyzing big data on the life of drilling tools. The present invention mainly analyzes and predicts the life of drilling tools through Bayesian inference, Markov process analysis, and artificial neural network algorithm (ANN), ensuring that high-quality tools that meet technical requirements can be provided on time for drilling engineering construction; by collecting and collating drilling tool data, using advanced computer algorithms for analysis, and predicting and analyzing the life of drilling tools, the system efficiency can be improved while reducing engineering costs, minimizing drilling risks, and improving the reliability and stability of the drilling system.
[0005] The purpose of the present invention is achieved through the following technical solutions.
[0006] A drilling tool life big data prediction and analysis method is carried out according to the following steps:
[0007] Step 1: Use the existing drilling reports to build a database, use a fully interconnected feedforward hidden layer network, i.e., back propagation learning rules, and establish a big data prediction model for drilling tool life based on drilling parameters;
[0008] The model building process:
[0009] First, the data samples are input to facilitate the application of convolution operation. The input size is Ntw×Nft, where Ntw represents the time series size and nft is the number of selected features. The original features are usually obtained from multiple sensor measurements. Next, 4 convolutional layers are stacked in the network for feature extraction. The configuration of these 4 layers is the same as the FN filter used, with a filter size of FL×1 and a filter size of 3×1. Then, the 2D feature map is flattened and connected with a fully connected layer. Dropout is used on the previous feature map, i.e., the flattening layer, to alleviate overfitting. Finally, a neuron is attached to the end of this network for RUL estimation. All layers use tanh as the activation function, and Xavier normal initializer is used for weight initialization. In order to further improve the prognostic performance, the back propagation (BP) algorithm is used for fine-tuning and the parameters of the model are updated to minimize the training error.
[0010] Step 2, using the drilling parameter-based big data prediction model for drilling tool life established in step 1 to automatically track the parameters of drilling tools, fit the existing theoretical model with the existing field experience data, and predict the tool life based on reliable actual data;
[0011] Step 3: Automatically detect drilling tool defects and abnormal conditions through artificial neural network algorithms (ANN), provide early warning of potential failures, formulate inspection plans based on risks, and avoid unnecessary routine periodic inspections and maintenance;
[0012] Among them, the early warning assessment process is to select multiple data for normalization, and use a sliding window to generate training and test data. Construct a deep neural network, and determine the number of hidden layers, the number and length of convolution filters. CNN uses standardized training data as input, and the labeled RUL value of the training sample is used as the target output of the network. The loss function uses RMSE, and the optimization function uses Adam. For the first 200 from the beginning, the learning rate for rapid optimization is 0.001, and then a learning rate of 0.0001 is used for stable convergence. Finally, the test data sample is input into the trained network for RUL estimation and the prognosis accuracy is obtained.
[0013] In step 1, the existing drilling reports include daily drilling reports (DDRS), section completion reports (EORS) and well completion reports (EOWR).
[0014] In step 1, the number of networks with fully interconnected feed-forward hidden layers is 2-3.
[0015] In step 1, the drilling parameters include: drilling pressure, rotation speed, well cleaning cleanliness, and drilling speed.
[0016] In step 2, the parameters of the drilling tool include: outer diameter, inner diameter, buckle type, length, wall thickness, specification, and model. The parameters of the drilling tool can be adjusted manually.
[0017] In step 2, the existing field experience data includes the allowable value of drill collar outer diameter wear and the allowable value of drill collar bending.
[0018] The allowable values of drill collar outer diameter wear are as follows:
[0019]
[0020] The allowable bending values of drill collar are as follows:
[0021]
[0022] Identification of connector type:
[0023] In the drilling process, it is very important to correctly select and connect the drilling tools. If the joints are mismatched, drilling accidents will occur. How to identify joints of different sizes and buckles in actual work? Generally, you should first look at the marking groove on the joint body. The positive buckle joint has a marking groove with a width of 10 mm and a depth of 1-1.5 mm; the reverse buckle joint has two marking grooves, the first groove is 10 mm wide and 1-1.5 mm deep, and the second groove is 5 mm wide and 1-1.5 mm deep. The words "421" and "620" are printed in steel code in the marking groove for identifying the buckle type of the joint. API drill pipe joints generally have no marking grooves, but they are also printed with size type codes. Experienced workers can also directly identify the size and type of the joint by observing the width of the buckle tip and the size of the taper of the rough buckle.
[0024] Commonly used API standard joint identification methods are as follows:
[0025]
[0026] IF-Internal Flat: FH-Internal Eye: REG-Regular
[0027] The beneficial effects of the present invention are as follows: the method can predict the life of drilling tools in advance, reduce tool failures during the drilling process, thereby improving work efficiency and achieving high production efficiency; use big data for information prediction to obtain the approximate life of tools under different conditions, thereby improving the efficiency of purchasing and material management. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 is the grayscale feature value of drilling tools;
[0029] Figure 2 It is the ACF and PACF graph of the original data;
[0030] Figure 3is the residual autocorrelation plot;
[0031] Figure 4 is the sample original data and the predicted results;
[0032] For ordinary technicians in this field, other relevant drawings can be obtained based on the above drawings without any creative work. DETAILED DESCRIPTION
[0033] The technical solution of the present invention is further described below through specific embodiments.
[0034] A drilling tool life big data prediction and analysis method is carried out according to the following steps:
[0035] Step 1: Use the existing drilling reports to build a database. The existing drilling reports include daily drilling reports (DDRS), well completion reports (EORS) and well completion reports (EOWR). Use a network with fully interconnected feedforward hidden layers. The number of fully interconnected feedforward hidden layers is 2-3, that is, the back propagation learning rule. A big data prediction model for drilling tool life is established based on drilling parameters. The drilling parameters include: drilling pressure, rotation speed, well cleaning cleanliness, and drilling speed.
[0036] The model building process:
[0037] First, the data samples are input to facilitate the application of convolution operation. The input size is Ntw×Nft, where Ntw represents the time series size and nft is the number of selected features. The original features are usually obtained from multiple sensor measurements. Next, 4 convolutional layers are stacked in the network for feature extraction. The configuration of these 4 layers is the same as the FN filter used, with a filter size of FL×1 and a filter size of 3×1. Then, the 2D feature map is flattened and connected with a fully connected layer. Dropout is used on the previous feature map, i.e., the flattening layer, to alleviate overfitting. Finally, a neuron is attached to the end of this network for RUL estimation. All layers use tanh as the activation function, and Xavier normal initializer is used for weight initialization. In order to further improve the prognostic performance, the back propagation (BP) algorithm is used for fine-tuning and the parameters of the model are updated to minimize the training error.
[0038] Step 2, using the drilling tool life big data prediction model based on drilling parameters established in step 1 to automatically track the parameters of interest to the drilling tool, the parameters of interest to the drilling tool include: outer diameter, inner diameter, buckle type, length, wall thickness, specification, model, the parameters of interest to the drilling tool can be manually adjusted, the existing theoretical model and the existing field experience data are fitted, the existing field experience data include the allowable wear value of the drill collar outer diameter and the allowable bending value of the drill collar, and the tool life is predicted based on reliable actual data;
[0039] Step 3: Automatically detect drilling tool defects and abnormal conditions through artificial neural network algorithms (ANN), provide early warning of potential failures, formulate inspection plans based on risks, and avoid unnecessary routine periodic inspections and maintenance;
[0040] Among them, the early warning assessment process is to select multiple data for normalization, and use a sliding window to generate training and test data. Construct a deep neural network, and determine the number of hidden layers, the number and length of convolution filters. CNN uses standardized training data as input, and the labeled RUL value of the training sample is used as the target output of the network. The loss function uses RMSE, and the optimization function uses Adam. For the first 200 from the beginning, the learning rate for rapid optimization is 0.001, and then a learning rate of 0.0001 is used for stable convergence. Finally, the test data sample is input into the trained network for RUL estimation and the prognosis accuracy is obtained.
[0041] The autoregressive difference moving average model in the data mining method first used the prediction principle of the model, then explained its prediction process, and proposed three indicators for evaluating the prediction model to describe the prediction situation of the model. An example study was conducted using the prediction model, and the prediction results of the model were obtained through example calculations. The prediction results were evaluated using MAPE, MAE and RMSE, and finally grey correlation analysis was used to make life predictions for the prediction results.
[0042] The autoregressive difference moving average (ARIMA) model, long short-term memory network (LSTM) and grey prediction model (GM) were studied respectively, and the oil field data were used for example analysis and result evaluation. Using two eigenvalues as prediction variables, the three types of prediction models were analyzed for example, which provided a reference for the selection of practical engineering applications.
[0043] The prediction results of several prediction models were compared and analyzed. Through the comparative analysis of the errors of several prediction results, it was found that the prediction accuracy based on LSTM was higher when grayscale features and Freeman chain code feature values were used as prediction variables.
[0044] For ease of explanation, spatial relative terms such as "upper", "lower", "left", "right" and the like are used in the embodiments to illustrate the relationship of one element or feature shown in the figure relative to another element or feature. It should be understood that, in addition to the orientation shown in the figure, the spatial terms are intended to include different orientations of the device in use or operation. For example, if the device in the figure is inverted, the element described as being "under" other elements or features will be positioned "on" other elements or features. Therefore, the exemplary term "under" can include both upper and lower orientations. The device can be positioned in other ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used here can be interpreted accordingly.
[0045] Furthermore, relational terms such as “first” and “second” and the like are merely used to distinguish one component from another having the same name, but do not necessarily require or imply any such actual relationship or order between these components.
[0046] The present invention is described in detail above, but the contents are only preferred embodiments of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of application of the present invention should still fall within the scope of the patent coverage of the present invention.
Claims
1. A drilling tool life big data prediction and analysis method, characterized by: Follow the steps below: Step 1: Use the existing drilling reports to build a database, use a fully interconnected feedforward hidden layer network, i.e., back propagation learning rules, and establish a big data prediction model for drilling tool life based on drilling parameters; The model building process: First, the data sample is input to facilitate the application of the convolution operation. The input size is Ntw×Nft, where Ntw represents the time series size and nft is the number of selected features. The original features are usually obtained from multiple sensor measurements. Next, 4 convolutional layers are stacked in the network for feature extraction. The configuration of these 4 layers is the same as the FN filter used, with a filter size of FL×1 and a filter size of 3×1. Then, the 2D feature map is flattened and connected with a fully connected layer. Dropout is used on the previous feature map, i.e., the flattening layer, to alleviate overfitting. Finally, a neuron is attached to the end of the network for RUL estimation. All layers use tanh as the activation function, and the Xavier normal initializer is used for weight initialization. In order to further improve the prognostic performance, the back propagation (BP) algorithm is used for fine-tuning, and the parameters of the model are updated to minimize the training error. Step 2, using the drilling parameter-based big data prediction model for drilling tool life established in step 1 to automatically track the parameters of drilling tools, fit the existing theoretical model with the existing field experience data, and predict the tool life based on reliable actual data; Step 3: Automatically detect drilling tool defects and abnormal conditions through artificial neural network algorithms (ANN), provide early warning of potential failures, formulate inspection plans based on risks, and avoid unnecessary routine periodic inspection and maintenance; Among them, the early warning assessment process is to select multiple data for normalization, use sliding windows to generate training and test data, build a deep neural network, and determine the number of hidden layers, the number and length of convolution filters. CNN uses standardized training data as input and uses the labeled RUL value of the training sample as the target output of the network. The loss function uses RMSE and the optimization function uses Adam. For the first 200 from the beginning, the learning rate for rapid optimization is 0.001, and then a learning rate of 0.0001 is used for stable convergence. Finally, the test data sample is input into the trained network for RUL estimation and the prognosis accuracy is obtained.
2. The method for predicting and analyzing drilling tool life big data according to claim 1, characterized in that: In step 1, the existing drilling reports include daily drilling report, DDRS, well section completion report, EORS, and well completion report, EOWR.
3. The method for predicting and analyzing drilling tool life big data according to claim 1, characterized in that: In step 1, the number of networks of fully interconnected feed-forward hidden layers is 2-3.
4. The method for predicting and analyzing drilling tool life big data according to claim 1, characterized in that: In step 1, the drilling parameters include: drilling pressure, rotation speed, well cleaning cleanliness, and drilling speed.
5. The method for predicting and analyzing drilling tool life big data according to claim 1, characterized in that: In step 2, the parameters of the drilling tool include: outer diameter, inner diameter, buckle type, length, wall thickness, specification, and model. The parameters of the drilling tool can be adjusted manually.
6. The method for predicting and analyzing drilling tool life big data according to claim 1, characterized in that: In step 2, the existing field experience data includes the allowable value of drill collar outer diameter wear and the allowable value of drill collar bending.
