KR102694136B1 - Oil gas plant equipment failure prediction and diagnosis system based on artificial intelligence - Google Patents

Abstract

The AI-based oil and gas plant facility failure prediction and diagnosis system according to the embodiment performs failure prediction and diagnosis of the plant facility through health judgment including the presence or absence of defects, uniformity, and adequacy of mechanical properties of the plant facility. Through the embodiment, an AI-based health prediction model of the core facility of the demonstration plant is implemented, and the AI prediction model is optimized through linkage with the virtualized operation and plant maintenance simulation system. In addition, by constructing and verifying simulation and actual plant operation data for failure prediction and diagnosis, the system efficiency and diagnosis accuracy can be improved.

Description

{OIL GAS PLANT EQUIPMENT FAILURE PREDICTION AND DIAGNOSIS SYSTEM BASED ON ARTIFICIAL INTELLIGENCE}

The present disclosure relates to an oil and gas plant facility failure prediction and diagnosis system, and more specifically, to an artificial intelligence-based oil and gas plant facility failure prediction and diagnosis system that enhances the operation and maintenance of a gas oil plant through the application of virtualization and artificial intelligence technologies, and provides a diagnostic assessment model for process risks and a dynamic risk assessment model.

Unless otherwise indicated herein, the materials described in this section are not prior art to the claims of this application and their inclusion in this section is not an admission that they are prior art.

Recently, the oil and gas plant sector has been seeing an increase in project orders due to rising international oil prices and increased energy demand, and the market size is expanding to the Middle East and other regions around the world. In addition, cutting-edge technologies such as artificial intelligence (AI) and ICT are being introduced to plant operations to improve the operational efficiency and safety of oil and gas plants. In particular, research is being actively conducted on technologies related to the development of data analysis prediction models and virtualization platform technologies based on AI, AI-based process efficiency, risk diagnosis and assessment technologies, and process facility monitoring portals, AI-based facility safety and soundness prediction diagnosis and assessment technologies, and AI-based virtualization operation platform verification technologies.

In the case of developing a data analysis prediction model based on artificial intelligence and virtualization platform technology, there is a case of building a sensor-based virtual plant in a previous study. However, it was only used to diagnose the current status, and it did not reach the level of utilizing AI-based prediction models and simulation models utilizing big data for predictive diagnosis.

In addition, AI-based process efficiency, risk diagnosis and assessment technology, and process facility monitoring portal technology exist as commercial products that utilize 3D model lightweighting and simplification technology for virtualization, but all of them have limited support scope. In particular, in the case of AI-based facility safety and soundness prediction diagnosis and assessment technology development, despite the efforts of individual domestic plant companies to develop smart solutions, most of the technologies are at the original technology stage, so it is necessary to continuously overcome the conservatism in plant operation through verification.

Meanwhile, dynamic risk assessment technology is considered a new technology overseas, and the two types of areas that dynamic risk assessment should cover are risk assessment for changes during the process life cycle and tracking of worker activities (maintenance work). Complex data generated from various structures are collected, but the methods for processing and analyzing the information so that users can efficiently utilize the information are limited. In addition, the accuracy of diagnosis and prediction and the level of satisfaction accepted by users have not been empirically evaluated and known, and it is difficult to understand in advance how the automatic diagnosis and prediction method of artificial intelligence is performed for various scenario situations, so a simulator is needed. In addition, there is currently no quantitative dataset that can measure and compare the stability of plant facilities, and there is a lack of datasets related to predictive diagnosis.

In Korea, smart plants are being applied mainly in the petrochemical industry, but service solutions are using products from advanced foreign companies, and smart core equipment also has relatively low stability and reliability. In order to secure technology, sharing of real-time operation data and application of process equipment are required, but it is difficult to accumulate experience due to reasons such as corporate secrets and stable operation of plants. In addition, the domestic engineering market is small, and the dependence on major basic solutions such as foreign design and analysis that have monopolized the market is deepening, so the commercialization of related research and development results is not being achieved.

1. Korean Patent Registration No. 10-1199290 (2012.11.02) 2. Korean Patent Registration No. 10-2145323 (2020.08.11)

The artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to the embodiment performs failure prediction and diagnosis of plant equipment through health assessment including the presence or absence of defects in the plant equipment, uniformity, and adequacy of mechanical properties.

Through examples, we implement an AI-based health prediction model for core facilities of a demonstration plant, and optimize the AI prediction model through linking with a virtualized operation and plant maintenance simulation system. In addition, we build and verify simulations for fault prediction and diagnosis and actual plant operation data to improve system efficiency and diagnostic accuracy.

An artificial intelligence-based oil and gas plant facility failure prediction and diagnosis system according to an embodiment includes a simulation system construction module that designs and verifies a simulation system based on plant core facility hardware; an evaluation module that performs safety and soundness evaluation of oil and gas plant facilities using the simulation system; an integrated control module that performs plant integrated control based on artificial intelligence; a facility soundness judgment module that performs plant facility soundness prediction and diagnosis based on artificial intelligence; and a data construction module that constructs a dataset related to plant core facility equipment and external loads.

The artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system as described above establishes a safe health inspection system centered on constant monitoring and operation management, thereby reducing the possibility of continuous risk exposure of related personnel due to rapid growth in the plant maintenance and inspection field. In addition, it enables preemptive standardization of maintenance and health evaluation platforms through the implementation of diagnostic modules and platform software.

In addition, through examples, it is possible to verify plant operation and secure performance, and it is possible to achieve not only performance verification and experience accumulation of technology, but also reduction of government budget. In addition, it is possible to secure predictive diagnosis technology using domestic technology rather than predictive diagnosis that relies on foreign products, and to support the application of the latest technology to the plant market and efficient plant operation through real-time control monitoring and introduction of AI. In addition, through examples, it is possible to secure the technological competitiveness of the plant industry such as minimizing damage to life, environment, and property in various accident scenarios, and expanding overseas orders by securing reliability in the plant safety field. In addition, it is possible to create high added value and new businesses for the entire plant and various factories by securing predictive diagnosis source technology.

In addition, by considering both safety and cost in the operation of the system through examples, early detection of system defects, diagnosis of the type and severity of defects that have occurred, and prediction of the time of failure in advance through Prognostics and Health Management (PHM) is performed, thereby reducing the possibility that equipment defects may lead to large-scale accidents. In addition, by realizing condition-based maintenance that performs maintenance only when necessary, breaking away from the traditional periodic maintenance or preventive maintenance method, it is possible to create significant economic benefits.

The effects of the present invention are not limited to the effects described above, and should be understood to include all effects that can be inferred from the detailed description of the present invention or the composition of the invention described in the claims.

Figure 1 is a drawing showing the core operation and maintenance technologies according to the entire plant cycle.
Figure 2 is a diagram showing the data processing configuration of an artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to an embodiment.
Figure 3 is a drawing for explaining the process of implementing an artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to an embodiment.
Figure 4 is a diagram showing an example of an artificial neural network data sheet for data acquisition.

The advantages and features of the present invention, and the methods for achieving them, will become clear with reference to the embodiments described in detail below together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and the present embodiments are provided only to make the disclosure of the present invention complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

In describing embodiments of the present invention, if it is judged that a specific description of a known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description will be omitted. In addition, the terms described below are terms defined in consideration of functions in embodiments of the present invention, and these may vary depending on the intention or custom of the user or operator. Therefore, the definitions should be made based on the contents throughout this specification.

Figure 1 is a diagram showing the core operation and maintenance technologies according to the entire plant cycle.

The AI-based oil and gas plant facility failure prediction and diagnosis system according to the embodiment provides the safety, soundness diagnosis function and predictive diagnosis function of the plant, thereby predicting and diagnosing the failure of the plant facility through the soundness judgment including the presence or absence of defects, uniformity, and adequacy of mechanical properties of the plant facility. In addition, through the embodiment, an AI-based soundness prediction model of the core facility of the demonstration plant is implemented, and the AI prediction model is optimized through linkage with the virtualized operation and plant maintenance simulation system. In addition, by constructing and verifying simulation for failure prediction and diagnosis and actual plant operation data, it is possible to improve the system efficiency and diagnosis accuracy.

FIG. 2 is a diagram showing the data processing configuration of an artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to an embodiment.

Referring to FIG. 2, the artificial intelligence-based oil and gas plant facility failure prediction and diagnosis system according to the embodiment may be configured to include a simulation system construction module (100), an evaluation module (200), an integrated control module (300), a facility health judgment module (400), and a data construction module (500). The term 'module' used in this specification should be interpreted as including software, hardware, or a combination thereof, depending on the context in which the term is used. For example, the software may be machine language, firmware, embedded code, and application software. As another example, the hardware may be a circuit, a processor, a computer, an integrated circuit, an integrated circuit core, a sensor, a MEMS (Micro-Electro-Mechanical System), a passive device, or a combination thereof.

The simulation system construction module (100) constructs a simulation system that designs and verifies a simulation system based on plant core equipment hardware. In an embodiment, the simulation system construction module (100) can design a simulation system operation scenario based on core equipment hardware and verify the simulation system through the scenario.

The evaluation module (200) uses a simulation system to perform safety and soundness evaluations of oil and gas plant facilities. In an embodiment, the evaluation module (200) performs soundness judgments including the presence or absence of defects, uniformity, and suitability of mechanical properties of plant facilities through a simulation system. In an embodiment, the evaluation module (200) can use vibration diagnosis data of plant equipment machinery for safety and soundness evaluation and safety precision diagnosis. In an embodiment, vibration diagnosis data is analyzed to evaluate plant safety and soundness. In an embodiment, vibration diagnosis data is interpreted through time domain analysis, frequency domain analysis, and component analysis, and safety and soundness evaluation results are calculated based on the analysis results.

In the embodiment, the time domain analysis of the evaluation module (200) includes a process of analyzing the size or temporal variation of the amplitude, the impulsiveness and symmetry of the waveform, etc. by using a time waveform, a probability density function of the amplitude, etc. In the embodiment, the frequency domain analysis of the evaluation module (200) includes a process of identifying which frequency components (spectrum) are included in the vibration by using an FFT (fast Fourier transform), etc. The spatial domain analysis is to identify which trajectory the center of the rotation axis (vibration) moves along in space by using a Lissajous figure, etc.

In addition, in the embodiment, the evaluation module (200) can perform safety and soundness evaluation using a method of using a primary effect parameter that indicates the performance or function of the equipment and a secondary effect parameter generated by the operation of the equipment. The primary effect parameter is a status parameter observed when the equipment performs its original purpose, and in the embodiment, the status of the equipment is monitored using the primary effect parameter. In the embodiment, performance monitoring and behavior monitoring can be performed.

Secondary effect parameters are secondary status parameters such as vibration, sound, and temperature that change due to the operation of the equipment. In the embodiment, parameter data is collected through a sensor for parameter diagnosis, and the collected parameter data is analyzed to enable early detection of plant abnormalities and accurate identification of the cause or location of the abnormality. In the embodiment, primary effect parameters are mainly used for simple diagnosis, and secondary effect parameters can be used for both simple and precise diagnosis.

The integrated control module (300) performs integrated plant control based on artificial intelligence. In the embodiment, the integrated control module (300) monitors plant facility data in conjunction with a virtualized operation and maintenance platform, and responds to changing monitoring data characteristics through an adaptive machine learning model. In addition, the integrated control module (300) is an integrated control module;

You can analyze control and simulation environments utilizing IoT.

In the embodiment, the integrated control module (300) performs prognostic and health management (PHM) that early detects system faults, diagnoses the type and severity of the faults that have occurred, and predicts the timing of failures in advance, taking both risk and cost into consideration in the operation of the system. Through prognostic and health management, the possibility of a large-scale accident is reduced, and condition-based maintenance (CBM) is realized, which performs maintenance only when necessary, breaking away from the traditional periodic or preventive maintenance (PM) method. The integrated control module (300) according to the embodiment processes data obtained through sensors mounted on the system (Signal processing), extracts characteristic factors representing the soundness of the current system from the data (Feature extraction), and performs fault diagnosis (Fault diagnosis) to identify the cause and severity of the current fault and fault prognosis (Failure prognosis) to predict the time of occurrence of the future fault. The purpose of the signal processing of the integrated control module (300) according to the embodiment is to leave information necessary for fault diagnosis and remove other information for signals mixed with various other signals and noise. To this end, the embodiment performs a process of discrete signal separation to remove periodic signals other than fault signals and signal enhancement to enhance fault signals having a small size. The embodiment extracts features using data subjected to specific signal separation and fault signal enhancement processing to more easily obtain fault-related information. In addition, the embodiment can extract fault-related information through a signal decomposition process to decompose the original signal into various components without going through the two steps described above.

According to an embodiment, among the decomposed components, there is data in which a fault signal is clearly identified, so that a fault feature can be extracted through the data. In addition, in the embodiment, linear prediction, self-adaptive noise cancellation (SANC), and time synchronous averaging (TSA) processes can be used to remove signals unrelated to a fault signal. Linear prediction is a method of linearly modeling a deterministic part of a signal using past data to estimate a future output value. Self-adaptive noise cancellation is a signal processing method that uses a primary signal and a reference signal as a filter to optimize the reference signal to be similar to a specific signal inherent in the primary signal. In the embodiment, the primary input is an original signal in which noise and a signal generated by equipment are mixed, and the reference input is a signal generated only by the plant equipment. In the embodiment, the filter parameter is calculated to obtain the error (e) by removing the noise signal in the basic input through the optimization process, so that only the desired plant signal can be selected. Time Synchronous Averaging is a method used to remove noise inherent in a signal and leave only signals showing periodic characteristics. In the embodiment, the vibration signal of the plant equipment is resampled to match the rotational period of the currently rotating shaft, and then the average value thereof is calculated. Through this, random noise components unrelated to the fault signal are removed, and only signal components having periodic properties according to the rotation remain. For this purpose, in the embodiment, a tachometer that generates a pulse signal for each rotation to measure the rotational period of the shaft or an optical encoder that measures the angle of the shaft can be used.

The facility health judgment module (400) performs AI-based plant facility health prediction and diagnosis. In the embodiment, health may include the presence or absence of defects in the plant facility, uniformity, suitability of mechanical properties, etc. In the embodiment, the facility health judgment module (400) can detect abnormalities by measuring signal data such as vibration or noise, and performs various signal processing such as frequency analysis as a data preprocessing process. In the embodiment, the facility health judgment module (400) can analyze vibration data to determine facility abnormalities. In the embodiment, a method of monitoring the presence or absence of an abnormality by measuring the Mahalanobis distance based on normal state data is used. In the embodiment, the Mahalanobis distance is measured for a vector composed of statistical indicators extracted from signal data for each section, and whether or not the corresponding section is in an abnormal state is determined. Mahalanobis distance is a distance measure frequently used in cluster analysis. When there is a correlation between variables in multivariate data, the distance is measured by standardizing data using a covariance matrix. In addition, in the embodiment, the Mahalanobis distance classifier pre-configures a Mahalanobis space from training data, and then measures the distance to the space for arbitrary data to determine whether it is in the same class as the training data based on the distance. In addition, in the embodiment, the performance of plant equipment abnormality detection can be improved through a preprocessing process of signal data. In the embodiment, the preprocessing process can include standardization, a Hamming window, cepstrum analysis, and statistical analysis.

In the embodiment, a Hamming window is applied to eliminate leakage errors that occur during the signal data collection process in the preprocessing stage, and cepstrum analysis is performed to separate formants, which are original signals of the signal data. The Hamming window and cepstrum analysis are performed to eliminate leakage errors, which are discontinuous amounts that occur during the process of arbitrarily segmenting continuous data. When the Hamming window is applied according to the embodiment, an accurate frequency domain spectrum can be obtained if the length of the extracted signal data is exactly an integer multiple of the signal period, but if not, an error occurs due to discontinuity at both ends. In the embodiment, a window function that converts signal data to values closer to 0 from the center to both ends is applied to eliminate this, so that a more accurate spectrum can be obtained from the FFT result. In the embodiment, a Hamming window is applied using Equation 1 to minimize this data leakage effect. In Equation 1, N represents the length of the window.

Mathematical Formula 1

Cepstrum analysis is a method of applying a fast Fourier transformation (FFT) to inversely Fourier transform (FFT) the logarithmic scale of a signal expressed in the frequency domain. The response to a general impulse signal on the time axis is expressed as a complex signal by convolving the input signal and the transfer function. When cepstrum analysis is applied, the waveform of the input signal and the transfer function that implements the response can be separated. In the embodiment, by applying cepstrum, the logarithmic function is separated into a function formed by two products and expressed as a sum, thereby separating the magnitude and phase of the signal converted to the frequency domain.

The data construction module (500) constructs a data set related to plant core equipment and external loads. In an embodiment, the data construction module (500) responds to changing data characteristics through an adaptive machine learning model, constructs a system integration data set after determining the final artificial intelligence model, and constructs a data set for application to oil/gas plant demand companies. In addition, in an embodiment, the data construction module (500) compares software figures and data set data through monitoring data measurement in oil and gas plant facilities, and constructs basic data for analyzing measured data and existing data.

Figure 3 is a drawing for explaining the process of implementing an artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to an embodiment.

Referring to FIG. 3, in the embodiment, a core facility hardware-based simulation system, a core facility hardware-based simulation system operation scenario, a facility integrated control platform utilizing IoT sensors, and a simulation system are designed. In addition, a facility integrated control platform and a safety soundness evaluation simulation system are implemented, and the simulation system is advanced through the development of AI-based soundness prediction diagnosis technology. To this end, sensor data for identifying the plant status are defined and collected, and a demonstration plant core facility AI-based soundness prediction model is developed. In addition, a predictive diagnosis inspection data set is constructed, and a demonstration plant core facility hardware-based system soundness prediction diagnosis system is constructed.

In addition, in order to design a simulation system based on the hardware of the core facilities of the demonstration plant, an operation scenario of the simulation system based on the hardware of the core facilities of the demonstration plant is developed, and an integrated facility control platform and simulation system utilizing IoT sensors are designed.

Figure 4 is a diagram showing an example of an artificial neural network data sheet for data acquisition.

The artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to the embodiment can predict and diagnose plant abnormalities by collecting and analyzing data using various data mining analysis techniques such as artificial neural networks, linear discriminant analysis, principal component analysis, Mahalanobis distance classifier, and logistic regression.

Through the artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system according to the embodiment, the existing virtualized smart plant technology is further developed to enhance the platform construction technology that integrates prediction and simulation models and existing distributed systems. In addition, it is possible to improve the process efficiency tailored to each plant through anomaly detection using artificial intelligence, and to provide a framework and platform that can be expanded and applied to various processes.

In addition, the practicality of the existing plant safety training education system can be improved, and technological competitiveness can be secured compared to advanced companies by securing new application technologies for safety training methods for large-scale plant facilities. In addition, a technology development system that leads to individual sensor-data center-plant management can be established through the development of maintenance and predictive diagnosis techniques specialized for plants. In particular, predictive diagnosis technology using domestic technology can be secured rather than predictive diagnosis that relies on foreign products through predictive diagnosis evaluation technology for plant facility safety and soundness, and the application of the latest technology to the plant market and efficient plant operation can be supported through real-time control monitoring and introduction of AI. In addition, through examples, it is possible to secure the technological competitiveness of the plant industry by minimizing damage to life, environment, and property for various accident scenarios and expanding overseas orders by securing reliability in the plant safety field. In addition, by securing predictive diagnosis source technology through safety and soundness predictive diagnosis technology, it is possible to create high added value and new businesses for the entire plant and various factories.

The disclosed content is merely an example, and various modifications and implementations can be made by a person skilled in the art without departing from the gist of the claims claimed in the patent, so the scope of protection of the disclosed content is not limited to the specific embodiments described above.

Claims (7)

In an artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system,
A simulation system construction module that designs and verifies a simulation system based on plant core equipment hardware;
An evaluation module that performs safety and soundness evaluation of oil and gas plant facilities using the above simulation system;
An integrated control module that performs integrated control of plant facilities based on artificial intelligence;
A facility health judgment module that performs artificial intelligence-based plant facility health prediction and diagnosis; and
Includes a data construction module that builds a data set related to plant core equipment and external loads;
The above simulation system construction module;
Design a simulation system operation scenario based on core facility hardware, and verify the simulation system through the scenario.
The above evaluation module;
Through the simulation system, the health of the plant equipment, including the presence or absence of defects, uniformity, and suitability of mechanical properties, is assessed.
The above evaluation module
The safety and soundness evaluation results are derived through time domain analysis, frequency domain analysis and component analysis of vibration diagnosis data, and the time domain analysis includes a process of analyzing the size or temporal variation of amplitude, and the impulsiveness and symmetry of the waveform using a probability density function of the time waveform and amplitude, and the frequency domain analysis includes a process of identifying which frequency components (spectrum) are included in the vibration using FFT (fast Fourier transform).
The above evaluation module
We perform spatial domain analysis to determine the spatial trajectory along which the center of the rotation axis moves using the Lissajous figure.
The above evaluation module
A safety and soundness evaluation is performed using the primary effect parameters that indicate the performance or function of the equipment and the secondary effect parameters that occur due to the operation of the equipment.
The above primary effect parameter is a status parameter observed in the facility, and the evaluation module monitors the status of the facility through performance monitoring and behavior monitoring using the primary effect parameter.
The above secondary effect parameters are the state parameters of vibration, sound, and temperature that change due to the operation of the equipment.
The above evaluation module collects parameter data through sensors for parameter diagnosis, analyzes the collected parameter data to diagnose early plant abnormalities and the cause or location of abnormalities, and the first effect parameter is used for simple diagnosis, and the second effect parameter is used for both simple diagnosis and precise diagnosis.
The above integrated control module;
In conjunction with the virtualized operation and maintenance platform, it monitors plant equipment data and responds to changing monitoring data characteristics through adaptive machine learning models.
The above integrated control module;
Analyze the control and simulation environment using IoT,
The above data construction module;
Respond to changing data characteristics through an adaptive machine learning model, build a system integration data set after deciding on the final artificial intelligence model, build a data set for application to plant demand companies, and compare data values monitored at oil and gas plant facilities with data set data to build basic data for analyzing monitored data and existing data.
Artificial intelligence-based oil and gas plant equipment failure prediction and diagnosis system.

delete delete delete delete delete delete