Risk computational modeling system and method for pipeline operation and integrity management

Abstract

A risk calculation modeling system and method for pipeline operation and integrity management is provided, wherein the system stores field data obtained from the pipeline operation system, online data obtained from the online inspection device, and external data from the camera on the external accompanying device of the online inspection device. The system processes the online data and field data to generate input data for risk modeling, and uses the input data to perform risk modeling on the pipeline to predict various failure risks of different parts and the entire pipeline, and supports the pipeline operation system and the online inspection device to implement risk monitoring on different locations.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] The priority claimed by this application under 35 U.S.C. §119 is based on U.S. Provisional Application No. 62 / 809,115, entitled “RISK COMPUTATIONAL MODELING SYSTEM AND METHOD FOR PIPELINES,” filed on February 22, 2019, which is hereby incorporated by reference and expressly incorporated herein.

[0003] The priority claimed by this application under 35 U.S.C. §119 is based on a continuation-in-part application of PCT patent application No. PCT / US2019 / 025438, entitled “Intelligent Data Acquisition System and Method for Pipeline”, filed on April 2, 2019, which is expressly incorporated herein by reference. The priority claimed by the PCT patent application under 35 U.S.C. §119 is based on U.S. Provisional Patent Application No. 62 / 651,520, entitled “Intelligent Data Acquisition System and Method for Pipeline”, filed on April 2, 2018, which is expressly incorporated herein by reference. Technical Field

[0004] The present invention relates to pipeline operations, and in particular to systems and methods for pipeline integrity and risk management. Background Art

[0005] Long-distance pipelines transport different products across cities, countries, and even across continents. In the United States, nearly 3 million miles of various pipelines transport nearly two-thirds of the country's energy needs. Whether in sparsely populated areas or densely populated areas, there are crisscrossing pipeline transportation networks. Pipeline transportation products include natural gas, crude oil, petroleum refineries, and other dangerous liquids. Pipeline transportation products are all flammable and explosive dangerous goods. Therefore, improving the safety and reliability of pipeline transportation can not only ensure the unobstructed "national economic artery", but also eliminate hidden dangers, reduce risks, and avoid serious damage or harm to people's lives and property, public facilities, life and the natural environment due to pipeline failures.

[0006] Pipeline risk models are a fundamental part of the risk assessment of pipeline operations. Risk models are at the heart of an operator's risk management system. Pipeline risk models and the results they produce are at the heart of many pipeline integrity programs in the United States and internationally. A risk model is a simplified representation of a pipeline system that represents the relationships between significant risk factors and a set of algorithms or rules that use available information and data relationships to perform risk assessments.

[0007] The pipeline industry began applying risk modeling to support risk assessment and decision making for planning maintenance and capital projects as early as the mid-1980s and has worked to develop industry standards and recommended practices such as API Recommended Practice (RP) 1160 - Managing System Integrity for Hazardous Liquid Pipelines and ASME B31.8S - Managing System Integrity for Natural Gas Pipelines.

[0008] Federal pipeline safety integrity management (IM) regulations in the United States require pipeline operators to use risk assessments. Risk models should support risk analysis, risk management decisions, and help operators evaluate and quantify the effects of various risk mitigation activities and options.

[0009] However, based on pipeline inspections and failure investigations, the U.S. Department of Transportation’s Pipeline and Hazardous Materials Safety Administration (PHMSA) and the National Transportation Safety Board (NTSB) have determined that it is imperative that pipeline operators perform risk assessments for their integrity management (IM) programs.

[0010] Operators should select the best model approach and then enrich the model with the best information about risk factors or threats for each pipeline segment and improve that data over time.

[0011] Currently, more than 90 risk assessment methods have been developed and applied. Risk models used in pipeline risk analysis can be classified based on the nature of the model inputs, outputs, and the nature of the algorithms used to convert the inputs to outputs. The following is a division of risk model categories:

[0012] Qualitative Model

[0013] Relative valuation or index model

[0014] Quantitative system model

[0015] Probabilistic Model

[0016] Based on the definition of the U.S. Department of Transportation's Pipeline and Hazardous Materials Safety Administration (PHMSA), the risk model describes the system risk of the entire pipeline by combining inputs related to the probability of failure of pipeline unexpected threats and their consequences of failure.

[0017] The conceptual definition used to constitute risk in risk assessment is given by the following formula:

[0018] Risk = Likelihood × Consequences

[0019] Risk is therefore defined as a measure of potential loss based on the likelihood (or frequency) of an event and the magnitude of its consequences.

[0020] For pipeline systems that transport hazardous liquids and natural gas, the most desirable event to avoid is a failure in the pipeline or any part of the pipeline system, and the resulting leakage of gas or hazardous liquids. The probability is the probability or frequency of failure due to threats affecting the pipeline, and the result is the severity of the impact on different receptor categories (e.g., life safety, environment, property) due to pipeline failure.

[0021] Risk analysis considers all potential and existing threats and the resulting failure potential for each segment along the pipeline. Therefore, risk modeling and analysis should include three key elements:

[0022] 1. Identify threats;

[0023] 2. Determine the probability of failure due to threats;

[0024] 3. Evaluate the consequences of pipeline failure.

[0025] Obviously, identifying threats is the cornerstone and foundation of risk modeling.

[0026] However, according to data from Rick Kowalewski’s report on “Project Assessment: Pipeline Integrity Management” (October 31, 2013), there were 1,355 major hazardous liquid pipeline accidents from 2002 to 2012, of which 824 were caused by corrosion and material failure, accounting for 60% of the total; the remaining 40% of the accidents were caused by excavation damage, human error, natural forces, etc., and of the 821 major gas pipeline accidents from 2002 to 2012, 420 of them, about 51%, were caused by corrosion and material failure, and the remaining 49% were caused by excavation damage, human error, natural forces, etc.

[0027] If we consider "corrosion and material failure" as threats related to time, there are still about 40 to 50% of major accidents caused by risk factors that are not related to time. In other words, external force damage, casualties, economic and environmental adverse effects all bring uncertainty and ambiguity to pipeline failure. Therefore, we need sufficient data, complete data, accurate data, and timely data to identify threats and find a mechanism to deal with unknown or uncertain risk factors that lead to pipeline failure.

[0028] Therefore, the system solution should include a comprehensive, thorough, more efficient and effective pipeline system inspection method, which requires technical updates and innovations. Then, a new generation of full-system solutions or well-designed frameworks based on technological innovation and data integration can overcome the challenges brought by insufficient or uncertain data in the real pipeline world, thereby solving the problems of risk ambiguity and uncertainty. Summary of the invention

[0029] In one aspect, a system for pipeline operation and integrity management, the system comprising at least one memory, including a non-transitory memory, for storing field data obtained from a pipeline operation system and online data obtained from an online inspection vehicle; at least one processor configured to perform data processing on the online data and the field data to generate input data for risk modeling; perform risk modeling of the pipeline using the input data to predict the risk of one of a plurality of failure mode states at the pipeline; and initiate risk monitoring of the portion of the pipeline through a pipeline operation system and an online detection device.

[0030] On the other hand, a system for pipeline operation and integrity management includes at least one memory, including one or more non-transitory memories, for storing field data obtained from an operating system, online data obtained from an online inspection vehicle, and external data from a camera on an external accompanying device; at least one processor, configured to generate a set of input value sets using field data obtained from the pipeline operation system, online data obtained from an online inspection vehicle, and external data from a camera on an external accompanying device; using risk modeling analysis to process the set of input values ​​to generate a set of output values, the set of output values ​​including a risk prediction of one of a plurality of failure mode states at the pipeline; and initiating risk monitoring of the portion of the pipeline through the pipeline operating system and the online detection device and its companion.

[0031] In one or more of the above aspects, the plurality of failure mode states include leakage, small leakage, large leakage, rupture and breakdown.

[0032] In one or more of the above aspects, the at least one processor is configured to perform risk modeling on the pipeline in the following manner: using a cluster machine that defines multiple sets of transfer functions to process inputs, wherein the multiple sets of transfer functions include a set of internal transfer functions having at least three subsets of corrosion-based transfer functions, stress-based transfer functions, and pressure-defect-based transfer functions.

[0033] In one or more of the above aspects, the cluster machine also includes a cluster of multiple events defined, wherein the cluster of multiple events includes at least: a cluster of risk events related to operating operations, a cluster of random risk events, and a cluster of risk events related to time.

[0034] In one or more of the above aspects, the cluster machine M is defined as:

[0035] M= <X,S,Y,δ i ,δ e ,δ c,λ,T,O,R,ta>, where X is the input value set, S is the sequence state set, Y is a set of output values, δ i is a set of internal transfer functions, δ e is a set of external transfer functions, δ c is a set of convergent transfer functions, T is the time-varying risk event cluster, O is the risk event cluster related to operation, R is the random risk event cluster, λ is the output function set, and ta is the time advancement function.

[0036] In one or more of the above aspects, the at least one processor is configured to define the combined set of transfer functions to include a subset of corrosion-based transfer functions, a subset of stress-based transfer functions, and a subset of pressure-defect based transfer functions.

[0037] In one or more of the above aspects, the at least one processor is also configured to receive and store in a memory the geographic data obtained from an external companion device for the in-pipe online detection device, wherein the companion is outside the pipeline and is configured to determine the geographic data of the online inspection device and generate information transmitted to the intelligent gateway, including geographic data, and video data received from a camera in the companion device and stored in the memory.

[0038] In one or more of the above aspects, the at least one processor is further configured to correlate the geographic data with online data from an in-pipe online inspection device to determine a location of the online data along the length of the pipeline.

[0039] In one or more of the above aspects, the at least one processor is further configured to perform data processing on pipeline online data, field data, and pipeline external data to generate input data for risk modeling.

[0040] In one or more of the above aspects, the pipeline internal online inspection equipment includes an excitation and related devices, and the excitation and related devices are configured to generate a magnetic field signal in the pipeline.

[0041] In one or more of the above aspects, the companion device outside the pipeline is configured to detect magnetic field signals from online inspection equipment inside the pipeline; determine geographic data including location information of the online inspection equipment inside the pipe; and generate relevant information containing geographic data and transmit it to the intelligent gateway.

[0042] In one or more of the above aspects, the in-pipe online inspection device includes an antenna configured to interact with the pipe wall to detect a very low or ultra low frequency signal from the smart gateway, wherein the very low or ultra low frequency signal also includes geographic data.

[0043] In one or more of the above aspects, the in-pipe online inspection equipment includes a plurality of pressure sensors, and the plurality of pressure sensors are arranged at different positions around the in-pipe inspection vehicle and near the circumference of the inner wall of the pipeline.

[0044] In one or more of the above aspects, the in-pipeline online inspection device is configured to determine a leak in the pipeline based on a difference between two or more pressure readings from the plurality of pressure sensors being greater than a predetermined threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1A An exemplary embodiment of an operational system for pipeline transportation is shown.

[0046] Figure 1B An exemplary embodiment of an intelligent integrated detection system for pipelines is shown.

[0047] Figure 1C An exemplary embodiment of an output data set of an intelligent integrated detection system is shown.

[0048] Figure 2A An exemplary embodiment of a finite state machine for pipeline risk modeling having multiple states is shown.

[0049] Figure 2B An exemplary embodiment of a cluster machine having inputs and outputs is shown.

[0050] Figure 3 An exemplary embodiment of a process for supporting a risk computational modeling system for operational operations and integrity management is shown.

[0051] Figure 4 An exemplary embodiment of the structure of a risk computational modeling system is shown.

[0052] Figure 5A An exemplary embodiment of a phased dynamic risk modeling life cycle is shown.

[0053] Figure 5B An exemplary embodiment of a workflow diagram of a risk computational modeling system is shown.

[0054] Fig. 6A An exemplary embodiment of an intelligent integrated detection system and the cooperation of its inspection equipment is shown.

[0055] Figure 6B An exemplary embodiment of a cross section showing a pressure test for leak detection with an internal smart data acquisition device is shown.

[0056] Figure 7An exemplary embodiment of the structure and workflow of the central processing unit is shown.

[0057] Figure 8 An exemplary embodiment of a framework for predicting risk consequence severity levels is shown. DETAILED DESCRIPTION

[0058] The words "exemplary" or "embodiment" are used herein to mean "serving as an example, instance, or illustration." Any implementation or aspect described herein as "exemplary" or "embodiment" is not necessarily to be construed as preferred or advantageous over other aspects of the disclosure. Likewise, the term "aspect" does not require that all aspects disclosed herein include the discussed features, advantages, or modes of operation. Obviously, an automatically adjustable, self-propelled online inspection vehicle can address these challenges.

[0059] The implementation examples will now be described in detail with reference to the accompanying drawings. In the following description, many specific details are set forth in order to provide an understanding of the various aspects described herein. However, it will be apparent to those skilled in the art that these and other aspects may be practiced without some or all of these specific components or with alternative components. In addition, in order not to obscure the aspects disclosed herein, known steps in the process methods presented herein may be omitted. Similarly, well-known components in the device may be omitted from the drawings and descriptions given herein to avoid obscuring the various aspects of the present disclosure.

[0060] The risk calculation modeling system demonstrated in this embodiment is a multi-level, multi-stage, data-oriented, event-driven, dynamic cycle full-system pipeline solution based on technical updates and innovations in pipeline detection methods. This embodiment also includes the establishment of an intelligent integrated inspection mechanism, a cluster machine with an 11-tuple structure, dynamic event-driven risk modeling for pipelines, data mining, machine learning and AI technology, and discrete event system specification (DEVS) simulation.

[0061] Figure 1AAn exemplary embodiment of an operational system for pipeline transportation is shown. In this example, the operational system includes a pipeline supervisory control and data acquisition (SCADA) system 130. The SCADA system 130 monitors and controls pipeline transportation. For example, the SCADA system can monitor and control field instruments 134, including flow, pressure and temperature meters, along the outer length of the pipeline. These field instruments are installed at certain specific locations along the length of the pipeline, such as injection or transmission stations, pump stations (liquid pipelines) or compressor stations (natural gas pipelines) and shut-off valve stations. The information measured by these field instruments is then collected in a local remote terminal unit (RTU) 131, which transmits the field data in real time to a central SCADA system server 132 using a communication system such as a satellite channel, microwave or cellular network. Therefore, the SCADA system 130 provides some real-time field data at discrete points along the length of the pipeline.

[0062] The SCADA system 130 can also remotely control and operate pipeline operations from what is commonly referred to as a "master control room." At this center, field data is consolidated into a central database. Data is received from multiple RTUs 131 along the pipeline. It is common to find RTUs installed at each station along the pipeline. The SCADA system in the master control room receives field data and displays it to the pipeline operator through a set of screens or human-machine interfaces to show the operating status of the pipeline. For example, an operator can monitor the hydraulic conditions of the pipeline and send operational commands to the field through the SCADA system (open / close valves, turn on / off compressors or pumps, change set points, etc.).

[0063] However, the SCADA system 130 is not ideal for leak detection and risk modeling. Pinhole leaks that release only a small amount (<1.5% flow) may not be detected by the SCADA system. If not discovered, such pinhole leaks may accumulate into large leaks. Even if the pipeline leakage amount is within the SCADA detection limit, sometimes the pipeline leak is misunderstood by the pipeline operator as a pump failure or other problem. Therefore, small leaks, especially pinhole leaks, are a challenging problem for today's pipeline operators because they are difficult to detect and pose a huge threat to the environment and public safety. Typically, decisions for pipeline operations such as alarm event handling, emergency shutdown, pipeline leak detection, and batch tracking are mainly based on the operator's experience or whether there are advanced pipeline application tools with different modeling and limited prediction accuracy.

[0064] Traditionally, risk models provide a representation of the risk in the entire pipeline system by combining inputs related to the likelihood and consequences of an unexpected release from the pipeline. Typically, the choice of risk assessment method depends on the inspection information, knowledge and organization, and additional information collected during pipeline operations.

[0065] In the report “A Critical Review of Candidate Pipeline Risk Models,” Smitha Koduru, Riski Adianto, Jason Skow (published by C-FER Technologies, December 2016) conducted a survey of 17 pipeline operators and received 13 risk models. The main learnings from the survey were that quantitative risk models have not yet replaced qualitative models for risk-based index scoring and that there is a lack of recognized standards for quantitative risk assessment. The report states that after reviewing 34 engineering and technology databases covering more than 6,000 scientific journals, a total of 70 publications were reviewed, only eight publications described a quantitative risk model system solution, and the other 62 publications dealt with failure consequence or failure frequency approaches for different threats.

[0066] Quantitative risk analysis (QRA) appears to be the most common decision-making approach for pipeline integrity management operations. QRA essentially uses probability to quantify the likelihood of risk and measurable parameters to quantify the consequences:

[0067] P t =1-∏(1-P i )

[0068] Pt: Total failure probability of the pipeline

[0069] Pi: Probability of failure of pipe segment due to threat i

[0070] Risk analysis models can be broadly classified into structural reliability methods such as Monte Carlo simulation, probabilistic models such as fault tree method and Bayesian network, and fuzzy logic models. However, in the process of using quantitative risk analysis (QRA), various uncertainties are inevitably introduced when estimating the probability of pipeline failure due to insufficient or uncertain data.

[0071] It is generally accepted that uncertainty remains in any model result, and that the input data is often a source of greater uncertainty than the computational methods. The solution with more data is likely to be the winner. Even a weaker model with less data is better than a more complex model with less data.

[0072] Therefore, general principles for designing robust next-generation pipeline risk models should include the following:

[0073] 1. Obtain more accurate data and additional data points to overcome the ambiguity of traditional risk modeling and methods.

[0074] 2. Identify all threats, including assessing potential threats and their interactions and how they increase risk, and identifying all risk drivers.

[0075] 3. Try to use the most complete and accurate reliable data, combined with root cause analysis of past events.

[0076] 4. Consider the non-correct operating factors, including the significant relevance of human-machine interaction and human performance to the possible failure or the significant impact on the consequences of the failure.

[0077] 5. Take into account random risk events (factors) such as weather, earthquakes, natural forces, and third-party threats such as excavation and damage.

[0078] 6. Consider a variety of scenarios to capture all possible broad-spectrum consequences, including outliers with severe consequences.

[0079] 7. Validate risk models based on incidents, leaks, historical failures and other historical information, ensuring the use of traceable and verifiable information and data.

[0080] The ideal situation is to achieve the above 7 basic principles by establishing an integrated inspection mechanism to merge the various major types of pipeline inspection methods into a more comprehensive, risk-based inspection method system that allows pipeline operators to apply various inspection data, field instrumentation data, pipeline characteristics, events and annual reports to continuously improve their pipeline integrity management programs.

[0081] PCT patent application number PCT / US2019 / 025438, entitled "Intelligent Data Collection System and Method for Pipeline," filed on April 2, 2019, is expressly incorporated herein by reference. The patent application describes an online detection tool and a real-time positioning system based on a special communication infrastructure and mechanism, which can obtain high-quality, high-precision continuous pressure, flow and temperature curves along the pipeline, and combine the marked geographic data along the way to effectively improve the prediction and diagnosis of pipeline failures.

[0082] The U.S. Provisional Patent Application No. 62 / 816,008, entitled “Adaptive System Structure and Method for Online Inspection Vehicle of Pipeline”, filed on March 8, 2019, and the Chinese Patent Application No. 201910938724.3, entitled “In-Pipeline Inspection Vehicle”, filed on September 30, 2019, are expressly incorporated by reference herein. The patent application describes an adaptive online pipeline inspection vehicle. The online inspection vehicle has an adaptive structure, automatic adjustment, autonomous driving function, and is equipped with a renewable energy system.

[0083] The U.S. Patent Application No. 62 / 816,008, entitled “Renewable Power System and Method for Pipeline Inspection Tool” filed on January 10, 2020, and the Chinese Patent Application No. 20191134967.6, entitled “Renewable Power System for Pipeline In-Pipeline Inspection Device” filed on December 19, 2019, are hereby expressly incorporated by reference. The pipeline renewable and rechargeable power system is used for the adaptive control and autonomous propulsion functional operation of the pipeline online inspection device.

[0084] Figure 1B An exemplary embodiment of an intelligent integrated detection system 100 for a pipeline is shown. This embodiment is similar to PCT patent application number PCT / US2019 / 025438, entitled "Intelligent Data Acquisition System and Method for Pipeline", filed on April 2, 2019, which is expressly incorporated by reference herein.

[0085] The intelligent integrated detection system 100 includes online detection equipment 111 (such as a robot detector or an intelligent PIG), an unmanned aircraft or an external accompanying device 121, a supervisory control and data acquisition (SCADA) system server 132, an RTU system 131 and an intelligent gateway 102. In the embodiment described, the intelligent integrated detection system 100 includes an intelligent data acquisition system of at least three different data channels, which provides real-time positioning based on a special communication infrastructure and mechanism.

[0086] As shown below:

[0087] A signal 122 sent from an external companion device 121, the signal carrying real-time video information and geographic data;

[0088] The low frequency signal generator 103 and the transceiver 113 generate a low frequency signal with geographic data 114;

[0089] A magnetic signal 112 emitted by an online detection device 111;

[0090] and a data communication protocol 133 between the intelligent gateway 102 and the RTU system 131 via a network connection.

[0091] The smart gateway 102 receives geolocation data from the companion device 121. The companion device 121 is a drone or other type of vehicle for tracking the online detection device 111. Since it is difficult for the online detection device 111 and the smart gateway 102 to communicate bidirectionally through the pipeline 101, one-way communication can be achieved through "earth mode" communication.

[0092] The online detection equipment 111 includes sensors to collect real-time online data, such as pressure, temperature and flow distribution along the pipeline line. The online detection equipment 111 can also carry various online detection devices for non-destructive testing (NDT) of pipelines, including magnetic flux leakage (MFL) testing, ultrasonic testing, electromagnetic acoustic transducer (EMAT), guided wave ultrasonic testing (GWUT) and pipeline radial detectors. The types of detection devices can detect leaks, deformations, ruptures, corrosion, defects, changes in pipe diameter thickness, dents or other defects. The online detection device 111 thus obtains sufficient and high-precision continuity and integrity data along the pipeline line.

[0093] In the example described, the companion device 121 is implemented in an unmanned vehicle (e.g., an unmanned aircraft), optionally, carried by a manned vehicle or other type of vehicle. For example, for an offshore pipeline, the companion device 121 can be set in a ship or an underwater vehicle. The companion device 121 is configured to report the location of the online inspection device 111 to the intelligent gateway 102. On the one hand, the companion device 121 can carry a wireless transceiver for communicating via a radio network, a cellular network, a satellite network, or other wireless network. The companion device 121 may also include a wired transceiver interface, such as a USB port or other type of wired connection, to communicate with one or more other devices via a LAN, MAN, and / or WAN when wirelessly connected to the unintelligent gateway 102.

[0094] The companion device 121 follows or tracks the online inspection device 111 in the pipeline 101. The companion device 121 is also equipped with a magnetic sensor 137 for detecting the magnetic field signal 112 from the online inspection device 111. The online inspection device 111 is equipped with an excitation device 138 in a magnetic flux leakage (MFL) module, for example. The excitation device 138 generates a magnetic field signal 112 in the pipeline 101 and transmits it through soil, air and / or water to communicate with the companion device 121.

[0095] The companion device 121 uses the magnetic field signal 112 to track the online detection device 111 in the pipeline 101 and determines its geographical location through GPS 135. The companion device 121 periodically transmits messages to report GPS coordinates and timestamps. Based on the timestamp and the geographical data information transmitted by the intelligent gateway 102, the online detection device 111 determines the geographical location, and then the geographical data can be associated with various test data. Therefore, the algorithm determines the geographical location of each point along the length of the pipeline for the data collected by the online detection device 111.

[0096] The online detection device 111 can sometimes adjust its speed and state for different inspection locations. In response, the accompanying device 121 detects the magnetic field signal 112 from the online detection device 111 and adjusts its speed or direction. Therefore, the accompanying device 121 is configured to track the online detection device 111 along the pipeline 101 by utilizing the signal 112 from the online detection device 111.

[0097] The accompanying device 121 can also be equipped with a camera 136 to perform external inspection of the pipeline 101 to detect any soil displacement, cracks, pipeline damage or risks from weather or natural forces and third-party threats, such as potential excavation damage. The accompanying device 121 can thus enhance the pipeline real-time monitoring and integrity system by performing additional monitoring at the same time.

[0098] The Ultra Low Frequency (ULF) is a frequency range of electromagnetic waves between 300 Hz and 3 kHz specified by the ITU. Many types of waves in the ULF band can be observed in the magnetosphere and on the ground. Communication through the ground using this conductive field is called "Earth Mode" communication. The communication was first used in World War I, and the signal transmission device was called a "power buzzer", an electromechanical device that could generate high voltage DC pulses at 700 Hz at a frequency of 700 Hz. The band can also be used for communication in mines because it can penetrate the earth. There are experimental reports on attempts at frequencies of 0.83-8.76kHz, which include the ULF and VLF (3-30kHz) bandwidths, with some success, demonstrating that underground pipes and cables can facilitate conduction in certain directions, so signal attenuation may be less than expected.

[0099] In the embodiment, the intelligent gateway 102 is configured with a low frequency signal generator 103 and a transceiver 113, and the device is configured to generate a low frequency (ULF or VLF) modulated ULF or VLF signal using geographic data from the companion device 121. The geographic data includes location information from the GPS of the companion device 121, such as latitude, longitude, altitude, etc. The intelligent gateway 102 sends the geographic data to the online detection device 111 via the "earth mode" communication signal 114.

[0100] One or more antennas are installed on the online detection device 111 to effectively interact with the inner wall of the pipeline and detect the "earth mode" communication signal 114. The communication signal 114 is processed in real time by the communicator and GPS positioning module in the online inspection device 111. The online detection device 111 can determine its geographic location in response to communication time delays and geographic data including latitude, longitude and altitude. During the stop state, the online inspection device 111 can most effectively and accurately determine its position to receive the "earth mode" communication signal 114. Therefore, the position of the online detection device 111 in the pipeline 101 for measurement can be determined. The online detection device 111 then associates its position with the measurement data of the pipeline. It is described in more detail in the PCT application number PCT / US2019 / 025438 intelligent data acquisition system cited herein.

[0101] In the embodiment, the online detection device 111 is configured with a renewable power system, which includes a pressure-based generator and / or a thermoelectric generator, and the Chinese patent application No. 20191134967.6, entitled "Renewable Power System for In-Pipeline Detection Device", filed on December 19, 2019, is expressly incorporated by reference. The U.S. patent application No. 16 / 739,459, entitled "Renewable Power System and Method for In-Pipeline Detection Device", filed on January 10, 2020, is also described in more detail.

[0102] The online inspection device 111 also includes an adaptive inspection vehicle, which has a self-adjusting carrier frame carrying positioning rollers and an inspection device bracket, and an adaptive drive turbine for automatically adjusting the drive speed. Chinese Patent Application No. 201910938724.3, entitled "In-Pipeline Inspection Vehicle", filed on September 30, 2019, is hereby expressly incorporated by reference. U.S. Provisional Patent Application No. 62 / 816,008, entitled "Adaptive System Structure and Method for Online Inspection Vehicle of Pipeline", filed on March 8, 2019, is hereby incorporated by reference.

[0103] In the described embodiment, the online detection device 111 is a supplement to the SCADA system because the data and measurements from the online detection device 111 are compatible with the existing pipeline supervisory control and data acquisition (SCADA) system or its operating system.

[0104] Figure 1C An exemplary embodiment of an output data set of an intelligent integrated detection system is shown.

[0105] In the described embodiment, the renewable power system 143 and the adaptive online inspection vehicle 142, the intelligent online detection equipment 141 are configured to have automatic adjustment, self-adaptation and self-propulsion features, and cruise in the pipeline to perform more intelligent, accurate, effective, efficient and complete inspection of the pipeline.

[0106] By working in collaboration with drones or other external accompanying equipment 121 and the intelligent gateway 102 in the intelligent data acquisition system 144, the intelligent integrated detection system 100 comprehensively coordinates various major types of pipeline detection methods, integrates multiple data acquisition channels and data sources, and collects a variety of large-scale location pipeline data or attribute (for example, pipeline outside and pipeline inside) data sets 151, 152, 153, greatly reducing data uncertainty and system ambiguity.

[0107] The regularization of the intelligent integrated inspection mechanism and the coordinated cooperation of the main types of detection technologies inside and outside the pipeline can also reduce the chance of pipeline failure caused by random risk events.

[0108] By utilizing the above innovations in pipeline inspection technologies and tools, multiple data sets from different channels and sources can be collected, fused, and integrated to create a big data foundation for pipeline risk modeling. For example, geographic data can be used to combine internal data from an in-pipe inspection vehicle 142 with external pipeline data from a camera on a companion device 121.

[0109] However, the interactive threat matrix and fault tree model of oil and gas pipelines indicate that the system is too complex to effectively use quantitative risk analysis models to model the various threats and risk drivers and their interactive superpositions.

[0110] Taking advantage of the computer's processing power could lead to significant advances in pipeline risk modeling, so more computer resources should be considered. Using artificial intelligence methods for risk computational modeling is a promising solution to the challenges of dealing with the complexity of risk modeling systems and handling very large data sets.

[0111] A Turing machine is a computational mathematical model that can be seen as the ultimate powerful logic machine for finite logical mathematical processes. A Turing machine can be formally defined as a seven-element ordered group M:

[0112] M=<Q,Γ,b,Σ,δ,q0,F>

[0113] Q is a non-empty finite set of states;

[0114] Г is a non-empty finite alphabet;

[0115] b∈Г is a blank character, and is the only character that is allowed to appear an unlimited number of times;

[0116] is a non-empty finite input alphabet that does not contain special whitespace characters

[0117] q0∈Q is the starting state;

[0118] is a final or accepted state;

[0119] is the transfer function, where L and R indicate whether the read / write head moves to the left or right.

[0120] The original tape contents eventually stop at state F and are accepted by M.

[0121] On the basis of Turing machine, automata theory and discrete mathematics of computer science were finally formed. Automata theory and its finite state machines have been used in the modeling of discrete event dynamic systems (DEDS). Discrete event dynamic systems (DEDS) are discrete state event-driven systems whose state evolution depends entirely on the occurrence of asynchronous discrete events over time. According to the characteristics of its discrete state space and event-driven state transition mechanism, the pipeline risk modeling system can be regarded as a discrete event dynamic system (DEDS).

[0122] Figure 2A An exemplary embodiment of a finite state machine for pipeline risk modeling having multiple states is shown.

[0123] In this example, seven states are described: two initial states: "Normal" 201 indicates that a location of the pipeline is in a normal state, and "Annular Seam" 202 indicates that a welded location of the pipeline may potentially have softening and brittleness or other risks. Two transition states "Corroded" 203 and "Cracked" 204 are threats or risk drivers. Any of the three failure states of leakage 205, small leakage 206, and large leakage 207 can be considered as the final state. This diagram shows us that there are a large number of possible interactions and combinations between states and events. Although Figure 2A It is not a complete diagram with all causal relationship matrices, but it is enough to correctly illustrate the state transition mechanism driven by risk events, and the drivers of risks, threats, failures, events and their correlation and interaction in pipeline operation. However, this finite state machine model obviously faces the problem of state and event explosion due to too many causal combinations and interactions.

[0124] The Discrete Event System Specification (DEVS) defined by Bernard P.Zeigle is a formalized model of discrete event systems used for component modeling and simulation of discrete event systems (DES). The hierarchical and modular features of DEVS are described in the book "Modeling and Simulation Theory" by Bernard P.Zeigle, Herbert Praehofer, Tag Gon Kim (Academic Press, 2000, 2nd edition), which can be used to model, design, analyze and simulate complex discrete event systems. Similar to the Turing machine, the Discrete Event System Specification (DEVS) is a structure with eight tuples, which is defined as:

[0125] M= <X,S,Y,δ int ,δ ext ,δ con ,λ,ta>

[0126] in:

[0127] X is the set of input values,

[0128] S is a set of states,

[0129] Y is a set of output values,

[0130] δ int :S->S is the internal transfer function,

[0131] δ ext ,:Q×X->S is the external transfer function,

[0132] δ con ,:Q×X->S is the convergence transfer function,

[0133] ta:S->R + 0,∞ ,and R + 0,∞ And R+0,∞ is the set of positive real numbers with 0 and ∞

[0134] (Q = {(s, e) | s∈S, 0≤e≤ta(s)} is the total state set,

[0135] e is the time elapsed since the last conversion)

[0136] λ:S->Y is the output function

[0137] Simulation of simulator models in the DEVS discrete event system formal model allows for quantitative analysis of the reliability and performance of different simulator alternative designs.

[0138] DEVS can model complex dynamic systems using the abstract concept of discrete events, which is very suitable for modeling communication protocols, part movement in automated manufacturing systems, or logic in process control systems. However, DEVS may not be fully suitable for pipeline risk system modeling because DEVS cannot meet the requirements of all scenarios and uncertainties.

[0139] As an extended version of the finite state machine and the DEVS discrete event system formal model, the cluster machine is a finite state machine driven by risk event clusters.

[0140] The cluster machine shown here has a structure of 11 tuples for modeling a pipeline dynamic risk system:

[0141] M= <X,S,Y,δ i ,δ e ,δ c ,λ,T,O,R,ta>

[0142] in:

[0143] X: is a set of input values

[0144] -X v ∈X, is a set of input variables and their input values,

[0145] -X c ∈X is a set of input parameters and their constant values, S: is a set of states

[0146] -S i :S i ∈S, is a set of initial states,

[0147] -S t :S t ∈S, is a set of transition states,

[0148] -S f :S f ∈S, is a set of failure states,

[0149] -s e :s e ∈S f ,s e Is a final state among a set of failure states

[0150] Y: is a set of output values,

[0151] δ i :S->S , is the set of internal transfer functions,

[0152] -δ ic :δ ic∈δ i, Based on the subset of transfer functions of corrosion,

[0153] -δ is :δ is ∈δ i, A subset of stress-based transfer functions,

[0154] -δ ip :δ ip ∈δ i, A subset of pressure-defect based transfer functions,

[0155] δ e :Q×X->S is a set of external transfer functions,

[0156] δ c :Q×X->S is a set of convergent transfer functions,

[0157] T: is a cluster of risk events related to time (time-varying),

[0158] O: is a cluster of risk events related to operational operations.

[0159] R: is a cluster of random risk events,

[0160] λ:S->Y is the set of output functions,

[0161] ta:S->R + 0,∞ , and R+0,∞ is a set of positive real numbers with 0 and ∞, ta is a time advancement function (Q={(s,e)|s∈S,0≤e≤ta(s)} is the total state set,

[0162] e is the time elapsed since the last conversion)

[0163] δ ＝ δ i ∪δ e ∪δ c, The total set of delta transfer functions,

[0164] T={(X t, δ t )|X t ∈X,δ t ∈δ},

[0165] O={(X o, δ o )|X o ∈X,δ o ∈δ},

[0166] R = {(X r, δ r )|X r ∈Xt ,δ r ∈δ}

[0167] The cluster machine in the embodiment includes three risk event clusters: T, O, and R. The risk event clusters can be mapped to internal, external, or combined conversion functions and corresponding input variables and sets of their values.

[0168] Figure 2B An exemplary embodiment of a cluster machine for modeling a pipeline dynamic risk system having an input 210 and an output 220 is shown. In the embodiment, the failure mode state set S in the output 220 of the cluster machine f Five failure mode states are included: leakage 221, small leakage 222, large leakage 223, rupture 224 and breakdown 225. Each failure mode state may become the final end state Se. Once the end state is determined, the system shutdown can be initiated. Although five failure mode states are shown in the embodiment, fewer or more failure mode states can be defined in other embodiments.

[0169] The most common definition of pipeline modes in the European Gas Pipeline Incident Database includes three levels:

[0170] Small leak - hole size less than 20 mm (or d / D <= 0.2);

[0171] Large leak - hole size is 20-80 mm (or d / D>0.2);

[0172] Rupture - hole size greater than 80 mm (or d / D ≈ 1);

[0173] (Where d / D: ratio of the size of the leak hole to the pipe diameter)

[0174] However, a research report submitted to PHMSA stated that there were a total of 464 failure incidents on land gas pipelines between 2010 and 2013, and the research showed that leakage: 30%, rupture: 38%, breakdown: 20%, and others: 12%. Breakdown was listed as an important failure mode.

[0175] Leakage 221 is a small and slow leakage state, which is much lighter than a small leak. For example, a small leak can be defined as a hole size less than 20mm-5mm, while a leak should be a hole size less than 5mm, and it can also be considered as a critical point of pipeline failure or a threshold point of pipeline rupture. Rupture 224 is definitely considered the most serious failure mode or should end state Se, which is the worst case of pipeline failure.

[0176] like Figure 2B As shown, the input state set 214 in the system input 210 includes an initial state S i215 and transition state S t 216. Initial state S i 215 may include, for example, normal, welded, and bent / buckled. t 216 may include, for example, corrosion, cracks, dents, defects, and deformations. More additional and / or alternative initial conditions S i 215 and transition state S t 216 can also be achieved here.

[0177] The system input 210 also includes three discrete event clusters: a time-related risk event cluster 211, an operation-related risk event cluster 212, and a random risk cluster 213. The time-related risk event cluster 211 may include, for example, internal corrosion, external corrosion, rupture, and fatigue. The operation-related risk event cluster 212 includes pressure, temperature, and medium / flow. The random risk cluster 213 may include, for example, incorrect operation, natural disaster threats (e.g., weather, earthquakes), and third-party threats (e.g., vandalism, accidental excavation damage). More additional and / or alternative risk events may be included in the event cluster. More additional and / or alternative event clusters may also be included in the system input 210.

[0178] The working principle of the cluster machine is the same as that of a computer. It determines the next state from the current state according to a set of rules called transition functions. Therefore, for the possible interactions, superpositions and combinations between various states and events, the cluster machine can enumerate, calculate and determine various possible transition paths (pipeline failure paths) from the initial state to the final state.

[0179] As part of artificial intelligence (AI), machine learning is a data science technology that allows processing equipment to use existing data to predict future behaviors, results, and trends. The predictions are based on learning processes and detection patterns. Because the intelligent integrated detection system 100 comprehensively coordinates various major types of pipeline detection methods, integrates multiple data acquisition channels and data sources to compile a wide range of data sets of pipeline data points or attributes at various locations, machine learning or AI is an ideal tool for studying and discovering different types of pipeline failure mechanisms. The learning of failure mechanisms is based on the established big data foundation and the possible transition paths from the initial state to the final state, which can be concluded from the simulation of possible interactions, superpositions, and combinations between states and events.

[0180] Figure 3An exemplary embodiment of a process for supporting a risk computational modeling system for operational and integrity management is shown. Because a Turing machine is a general example of a central processing unit (CPU) that controls data operations performed by a computer, the central processing unit 302 of the risk computational modeling system 300 is a specific implementation on the physical structure form and device of a cluster machine, although other risk modeling techniques can also be implemented here. Figure 7 The structure and processing process of the central processing unit 302 are described in more detail.

[0181] The risk calculation modeling system 300 integrates data sets from different data sources and channels as system inputs 301. The data sets may also include historical data, such as online data obtained from online inspection tools along the inside of the pipeline, such as leaks, deformations, defects, ruptures, corrosion, defects, pressures, temperatures, flow rates, thickness changes, dents, or other data. The online data is associated with geographic data from the companion device 121 for integration. Historical data may also include external data of the pipeline from the camera of the companion device 121. Operational data includes field data obtained from an operating system (e.g., SCADA system 130) at discrete points along the pipeline. Geographic Information System (GIS) data and engineering data may also be included in the input data set. System input 301 may include other data, such as known excavations, third-party risks, natural force risks, ground displacements, weather risks, or leaks.

[0182] The central processing unit 302 includes a risk modeling engine 321 having two supporting components referred to as a failure mechanism learning engine 322 and a simulator 323. The risk modeling engine 321 is implemented using one or more processing circuits 325 communicatively coupled to one or more memories 324. In one aspect, the memories 324 may include one or more non-transitory processor-readable memories storing instructions to be executed by the one or more processing circuits 325 utilizing one or more functions described therein.

[0183] The system output 303 of the central processing unit is a collection of threat and risk drivers and predictions associated with failure mode states. The results and findings in the system output 303 can support two devices such as leak detectors 304 and risk monitors 305 to prevent pipeline failures by improving the accuracy of predictions and diagnostics, thereby fusing two discrete systems such as pipeline operation and management systems 341 and pipeline integrity management systems 351.

[0184] For example, in actual application, historical data, operational data, and GIS and engineering data are processed to generate system inputs for the risk computational modeling central processing unit 302. The risk computational modeling central processing unit 302 performs risk analysis on the system input 301 to generate system outputs 303. The risk computational modeling central processing unit 302 can generate outputs including failure mode state predictions, such as the risk of leakage of a certain portion of the pipeline due to a failure threat. The pipeline operation system 341 can then initiate risk monitoring of the portion of the pipeline and request an online inspection vehicle to perform additional inspections on the portion of the pipeline.

[0185] Figure 4 An exemplary embodiment of the structure of the risk computational modeling system 400 is shown.

[0186] The risk calculation modeling system 400 is used for pipeline operation and integrity management and includes a multi-layer architecture based on system specifications and simplification. In one embodiment, the multi-layer architecture includes a data connection layer 410, an inspection interface layer 420, a central risk modeling layer 430, and a supervisory control and monitoring layer 440.

[0187] At the data connection layer 410, data from both the outside and the inside are integrated with the risk modeling system 400. Data from relevant industrial libraries and external resources such as documents and inspection rules of integrity management regulations are modularized and stored as data sets in the failure event data server 411 and the integrity management data server 412. Historical inspection data is stored on the historical data server 413, and the historical inspection data includes inspection data of the pipeline 101 from the intelligent integrated inspection system 100, such as data from the online inspection device 111 and the matching accompanying equipment 121.

[0188] The operational data includes operational data about the pipeline and is mostly generated by the SCADA operating system and stored on the operational data server 414 .

[0189] Data about pipeline materials, processing, manufacturing, and pipeline engineering construction is stored on the manufacturing and engineering data server 415. Data related to design requirements for structural and operating parameters, as well as geographic and ground information along pipeline routes and pipeline segments are also modularized and stored in the design & ground data server 416. Unstructured data can be archived or stored in a NoSQL database using a big data solution for data mining on a data server. Although different servers are shown here, multiple data sets can be stored on one or more data servers, for example, in different databases. One or more servers each include a storage device for storing data. More additional or alternative data sets can be defined and implemented in the data connection layer 410. The data connection layer 410 therefore includes data from multiple source sources.

[0190] At the inspection interface layer 420, data source connections and data channel component interactions are constructed. Inspection instructions from the inspection scheduler 421 are sent to inspection equipment, such as online inspection equipment 422. As components of the intelligent integrated inspection system 100, the online inspection equipment 422, the intelligent data acquisition system 423 and other data from the camera 424 or from the accompanying equipment 121, and their output data sets are interpreted, combined and stored in the historical data server 413.

[0191] At the central risk modeling layer 430, the risk modeling machine 431, simulator 432 and failure mechanism learning machine 433 act as the central processing unit 302. The data input includes both new detection data sets from the inspection interface layer 420 and historical data sets and failure event data sets from the data connection layer 410.

[0192] At the supervisory control and monitoring layer 440, risk monitors 441 and leak detectors 442 interact with real-time transient models (RTTM) on the SCADA system and operating systems 443 to prevent pipeline failures by receiving, processing and visualizing analytical predictions, findings and conclusions from the central risk modeling layer 430.

[0193] Figure 5A An exemplary embodiment of a staged dynamic risk modeling lifecycle 500 is shown.

[0194] First, the detection stage 510 is jointly driven by the multi-channel detection equipment in the intelligent integrated detection system 100. The real-time data from the online detection equipment 111 is mapped and labeled with the corresponding geographic information during the collection process, and the collected data is effectively interpreted as corresponding attributes to form a wide-ranging, multi-dimensional data set.

[0195] Second, the data processing stage 520 groups, integrates and manages the operation of data from different sources, which can be historical data or newly collected field data from the operating system 130, and classifies the data into different risk event-driven clusters for modeling needs.

[0196] Third, the risk modeling phase 530 is based on data grouping, data resampling and data mining techniques. The simulation using DEVS, graph theory and failure mechanism learning is based on the principle of machine learning, which will Figure 7 Risk modeling 530 generates risk predictions and recommended actions.

[0197] Fourth, using the results and predictions from the risk modeling phase 530, the operations and integrity management phase 540 initiates maintenance, repairs, risk monitoring, leak detection, and new inspection plans and schedules to start the next round of risk management cycle.

[0198] Therefore, the risk calculation modeling system is a multi-stage, multi-level, inspection-based, data-oriented, event-driven, and dynamic-cycle full-system solution.

[0199] Figure 5B An exemplary embodiment of a workflow diagram of a risk computational modeling system is shown, explaining the phased dynamic risk modeling life cycle 500 in more detail.

[0200] The detection phase 510 includes inspection request and plan scheduling 511, external accompanying equipment and video detection 512, intelligent online detection 513, intelligent data acquisition 514, stress testing 515 and data interpretation 516.

[0201] The data processing stage 520 includes pipeline data segmentation 521 , data alignment 522 , and threat identification 523 .

[0202] In the third phase, risk modeling 530 includes data mining 531, failure mechanism learning 532, risk modeling / prediction 533, simulation / verification and confirmation 534, and experiment / failure path testing 535. Figure 7 Provide further description and explanation.

[0203] In the fourth phase 540, the analysis results, insights and predictions about threats and risk drivers 507 from the risk modeling phase 530 are input into risk monitors 541, leak detectors 542, RTTM (real-time transient model) and SCADA systems 543 to support monitoring, control, operational monitoring of risk vulnerabilities and detection of leaks or possible serious failures before serious failures. The results about threats and risk drivers 507 are also transmitted to inspection requests and scheduling 511 for application of pipeline integrity management projects and to start the next round of cycles.

[0204] The analysis results, insights and predictions of threats and risk drivers 507 from the risk modeling stage 530 are transmitted as instructions for external or internal inspections to external companion devices and video detection 512, intelligent online detection 513, and intelligent data collection 514, triggering detailed inspections and data collection of some key locations and key threats. Insights such as pattern recognition in machine learning can be integrated with detection devices to make them more perceptive, sensitive and intelligent in problem identification.

[0205] Current regulations for both gas and liquid pipelines in pipeline integrity management require risk analysis and data integration assessment. In fact, this is the most common problem and challenge, because for various reasons, data is often collected in different formats and there is no common identifier, so effective integration is almost impossible. The collaborative work of different detection devices in the intelligent integrated detection system 100 can help overcome the challenges and difficulties faced by the pipeline industry.

[0206] Fig. 6A An exemplary embodiment of an intelligent integrated detection system 600 and its inspection equipment working in coordination is shown.

[0207] Basically, the sensors and test equipment for online detection 513 and intelligent data acquisition 514 in Figure 5 can be installed and carried on an adaptive detection vehicle to construct an intelligent integrated online detection device 611, which is equipped with a renewable and rechargeable power system for self-propulsion and adaptive control of online detection.

[0208] The online detection device 611 can adaptively navigate in the pipeline 601 and be guided by a companion device such as a drone 621. The communication structure between the devices is composed of a wireless signal 622, an intelligent gateway communicator 602, a low-frequency signal generator 603 and a transceiver 613, and a message signal 614 with geographic data.

[0209] The interaction between the online detection device 611 and the external detection device 621 works together to generate the pipeline internal detection data set of the online detection device 611 and the external data set of the external inspection device 621 and the common pipeline location between them. The internal and external data sets of the pipeline can be mapped and marked using common geographic information, including universal identifiers along the pipeline location during inspection and data collection. Such mapping and marking can improve data quality and enable effective data integration.

[0210] For example, at a location 604 with ground displacement, soil movement may cause pipe deformation 622. Both the in-pipe detection device 611 and the external detection device (e.g., a camera on a drone 621) can detect the pipe deformation 622 from different angles. The online detection device 611 and the drone 621 collect data at different detection orientation channels (inside and outside the pipe) and generate multiple data sets. Multiple data sets from internal and external measurements of the pipe related to the same deformation 604 can be integrated using the labeled geographic information and then processed for risk modeling and risk assessment.

[0211] About 40%-50% of major accidents are caused by time-independent events or random risk events, which brings huge uncertainty to the prediction of pipeline failure. For random risk events, such as third-party / mechanical damage, excavation, vandalism, weather-related factors and external forces, frequent external inspections by drones, and joint inspections by online detection equipment 611 and external inspection equipment 621 may detect potential problems faster and more timely, thereby reducing the chance of pipeline failure caused by random risk events to help overcome the uncertainty of system risks. In addition, the data of time-independent or random risk events collected by the intelligent integrated detection system 600 can be effectively used for risk modeling based on the cluster machine method to greatly improve the applicability and accuracy of the risk modeling system.

[0212] The ASMEB31.8S specification proposes that pressure testing can be used as one of the three methods for evaluating pipeline integrity. During the pressure testing process of this specification, the test pressure is 1.25 to 1.5 times the maximum allowable operating pressure (MAOP). It may take several days to fill the pipeline with water or other fluids / gases to obtain such a test pressure. The cost of pressure testing one mile of pipeline is between US$150,000 and US$500,000. Due to the low cost of online inspection, the cost per mile is only US$5,000, so in practice, online inspection tools are used to evaluate more than 92% of the total pipeline mileage. With its cost-effective advantage, the online detection equipment 611 of the intelligent integrated detection system 600 can perform the real-time pressure test 515 shown in Figure 5. The online detection device 611 can perform strength tests and leakage tests, Figure 6BA more detailed explanation is given.

[0213] Figure 6B An exemplary embodiment of a cross section of a pressure test for leak detection by an intelligent data acquisition device inside a pipeline is shown. In the described embodiment, 12 pressure sensors 6110, 6111, 6112, 613, 6114, 6115, 6116, 6117, 6118, 6119, 6120, 6121 are configured and installed on the internal detection device 611. The sensors are arranged around the online detection device 611 and placed near the inner wall of its pipeline 601. Under normal conditions, the pressure distribution along the inner wall of the pipeline 601 should be approximately uniform. Therefore, the pressure difference between any two pressure sensors is relatively small.

[0214] In the example, there is a leak point 6011 on the pipeline 601. Since the leak point 6011 is located between the pressure sensors 6111 and 6112, the pressure values ​​measured at the pressure sensors 6111 and 6112 should be lower than the pressure values ​​measured at the pressure sensors 6113 and 6110. The pressure difference between the sensors increases with the distance from the leak point 6011. Therefore, the maximum pressure difference should be between the pressure sensors 6112 and 6118 or between the pressure sensors 6111 and 6117. Compared with normal operating conditions, the pressure measurement difference between the two diagonal pressure sensors (for example, between 6111 and 6117 or between 6112 and 6118) is relatively large.

[0215] In use, the online detection device 611 collects pressure readings at multiple locations around the circumference of the inner wall of the pipeline 601 at approximately the same location along the length of the pipeline. When the difference between two or more pressure readings at different locations around the circumference is greater than a predetermined threshold, a leak is determined. Geographic information from the accompanying device 621 can provide the location of the leak point along the length of the pipeline. In addition, the difference between two or more pressure readings can indicate the size of the leak, such as a pinhole leak, a small leak, a large leak, or a rupture. Generally, the size of the leak increases as the difference between the two or more pressure readings increases. Therefore, a pressure stability test around the inner wall of the pipeline 601 can detect leaks and even identify the location of the leak along the length of the pipeline.

[0216] In theory, pressure loss occurs due to friction and the average velocity of the fluid along a given length of pipe. In fluid dynamics, the Darcy-Weisbach equation is an empirical model:

[0217]

[0218] The pressure loss per unit length Δp / L (unit: Pa / m) is a function of the following parameters and variables:

[0219] ρ, density of the fluid (kg / m3);

[0220] D, hydraulic diameter of the pipe (m);

[0221] ν, average flow velocity (m / s);

[0222] f D , Darcy friction coefficient (also called flow coefficient λ).

[0223] When the online detection device 611 moves axially along the pipeline, it measures the pressure values ​​at different positions to verify the empirical equation and help establish a new pressure equation model.

[0224] Pressure cycle data is used as a mechanism to predict fatigue crack growth. Pressure data collected by the online detection equipment 611 can also verify and modify such "working pressure cycle" models as follows:

[0225]

[0226] P X , the pressure at any point between stations

[0227] P1 = pressure generated at the upstream station during operation, psig

[0228] P2 = suction pressure of downstream station during operation, psig

[0229] K = psi / head foot

[0230] L1 = upstream miles, miles

[0231] L2 = Miles downstream, miles

[0232] L X = Miles from location X, miles

[0233] h1 = upstream elevation, feet

[0234] h2 = downstream elevation, feet

[0235] hx = elevation of location X, feet.

[0236] The method is used to determine the pressure circulation at selected locations between pumping stations.

[0237] The crack growth model can be used to evaluate the effect of pressure cycle induced growth on possible residual defects in the pipe. The crack growth rate caused by the pressure cycle spectrum is modeled using the Paris law equation:

[0238]

[0239] Where "a" is the crack depth,

[0240] "N" is the number of pressure cycles,

[0241] is the crack extension per cycle (da),

[0242] ΔK is the stress intensity factor for a given stress cycle.

[0243] The constant "C" and the exponent "n" represent the fatigue crack growth rate applicable to a particular material and environment.

[0244] Figure 1C 1 shows the output data set of the intelligent integrated detection system 100. The intelligent data acquisition device 111 carries multiple sensors, which are configured to detect one or more of the following: radial displacement, pressure, temperature, flow and acoustics to collect different data sets. Figure 5B As shown, intelligent data collection 514 and pressure testing 515 can establish sufficient and high-precision continuous data files along the pipeline route, thereby significantly improving the sensitivity, effectiveness and reliability of the RTTM / SCADA system 543 and the leak detector 542. Therefore, the system can effectively improve the probability of leak prediction and detection, and accurately determine the leak location, including tiny or pinhole leaks.

[0245] Based on the fusion of multiple data collection channels, data integration and real-time pressure testing, the output results and findings from the risk calculation modeling system can be used to support leak detection monitoring and risk monitoring to prevent pipeline failures and failures by improving the accuracy, sensitivity and reliability of pipeline operation system monitoring, thereby integrating pipeline operation and integrity management systems.

[0246] Figure 7 An exemplary embodiment of the structure and workflow of the central processing unit for the risk computational modeling system 400 is shown. The risk modeling engine 710 includes two supporting components, namely the simulator 720 and the failure mechanism learning engine 730.

[0247] The cluster machine (finite state machine with risk event clusters) is an 11-tuple structure used to model the risk dynamic system of pipelines. It defines three groups of functions (internal, external, and confluence) and three groups of risk events. The set of internal transfer functions is divided into three function subsets: corrosion-based, stress-based, and pressure-defect-based transfer functions.

[0248] Each risk event can be mapped into an internal, external or confluent transformation function and a corresponding set of input variables and their values. The Cluster Machine is different from the 8-tuple DEVS structure, but it can also be mapped into the DEVS formal structure, a set of atomic models and coupled models. Figure 7 The simulator 720 in the embodiment can be modeled in the form of a cluster machine or DEVS.

[0249] Experiment 721 can use a known set of internal conversion functions to simulate possible risks and failure states. The experimental results 721 are recorded as event cases and stored on the failure event data server 703. In fact, many pressure-defect-based functions have been proposed. The ASME B31G burst pressure equation for defective pipes is a model that can be applied in this risk modeling system:

[0250]

[0251] p: bursting pressure of defective pipeline, MPa;

[0252] when

[0253] when

[0254] σ=SMYS+68.95MPa as the flow stress of the pipeline, MPa;

[0255] D: pipe diameter, mm;

[0256] t: pipe wall thickness, mm;

[0257] d: defect depth, mm;

[0258] l: defect length, mm;

[0259] M: Risk factors

[0260] SMYS: Specified minimum yield stress

[0261] The most well-known transition functions for calculating residual strength in hydrocarbon transportation pipelines are the ASMEB31G and RSTRENG models:

[0262] ASME B31G Model:

[0263]

[0264]

[0265] RSTRENG model:

[0266]

[0267]

[0268] p is the failure pressure of the pipeline with active corrosion defects;

[0269] D: pipe diameter, mm;

[0270] t: pipe wall thickness, mm;

[0271] L: defect length;

[0272] y: defect depth;

[0273] YS: Yield Strength

[0274] There are more transfer functions for pipeline modeling related to corrosion, crack and pressure-defect based transfer functions. Multiple transfer functions with different risk cause events and their variable values ​​for experiment 721 can be implemented and modeled in the simulator 720, and modeled, and combined with the root event (cause) analysis of past events to generate more events and data cases.

[0275] Unlike traditional risk modeling methods such as interactive threat factor matrices, fault tree models for interactive threat factors, cluster machines can instantiate all state-based risk-failure paths and risk event-driven transfer functions. The summary of all possible paths can also be viewed as a fault-failure path tree, which is represented as a graph G:

[0276] G=(V,E,φ)

[0277] V: a set of vertices (also called nodes or points);

[0278] E: a set of edges (also called links or lines);

[0279] φ: E→{{x, y}|(x, y)∈V 2 ∧x≠y} A mapping function that maps each edge to an ordered pair of vertices

[0280] The failure path test 722 in the simulator 720 enumerates and simulates all possible paths derived from the cluster machines. This process also produces many useful results as failure event cases with a large data set.

[0281] As a supporting component of the risk modeling machine 710 and the simulator 720, the failure mechanism learning machine 730 learns the state transition mechanism driven by the risk event and determines the transition function defined in the cluster machine. The failure mechanism learning machine 730 is based on artificial intelligence and learning science. The learning science plays a key role in the fields of statistics, data mining and artificial intelligence. Therefore, the failure mechanism learning machine is a learning machine based on data mining and AI.

[0282] Data mining 731 is the process of discovering risk threat patterns and failure modes from huge data sets. These huge data sets include pipeline characteristics based on GPS location and geospatial, physical properties of soil, inspection and pipeline operation data from different data sources and collection channels. Consideration should be given to establishing thresholds for threats and pipeline failures associated with different risk events. Data mining methods such as cluster analysis, statistical classification, association rule learning should be used to prepare training data sets for failure mechanism learning 732. The risk threat patterns and failure modes identified in data mining 731 are grouped into different state subsets in the cluster machine.

[0283] Basically, by using a variety of machine learning methods and AI models, failure mechanism learning 732 learns the state transition mechanism driven by risk events and attempts to determine the transition function from the previous state to the next state in the cluster machine. The learning process 732 is based on the training data set in data mining 731, the data set in the fault failure event data server 703, and the operation data set in the operation data server 704.

[0284] Part of artificial intelligence (AI), machine learning is a data science technique that allows computers to use existing data to predict future behaviors, outcomes, and trends. These predictions are based on learning processes and detecting patterns. Supervised learning, which includes classification and regression, can be used to learn functions that map inputs to outputs based on a training data set. This is a method used to condition on having example input-output data pairs, which can be obtained from historical data sets. Unsupervised learning algorithms take a set of data containing only inputs and find structure in the data, such as groupings or clusters of data points.

[0285] According to the report from "Petroleum Pipeline Failure Prediction Models," regression analysis and artificial neural network (ANN) models were developed based on historical data of pipeline accidents. Both models were able to satisfactorily predict pipeline failures due to mechanical, operational, corrosion, third-party damage, and natural disasters, with an average effectiveness of 90% for the regression model and 92% for the ANN model.

[0286] The article “A Survey on the Application of Neural Networks in Oil and Gas Pipeline Safety Assessment” by M. Lavouni et al. (IEEE Symposium on Computer Intelligence in Engineering Solutions, 2014, pp. 95-102) found that neural networks have been widely used in the field of pipeline safety to predict the probability of failure, the causes of failure, the classification of metal defects, and the detection and sizing of metal defects on pipelines.

[0287] Bootstrap is an important statistical method for estimating the variance of a statistic and then performing interval estimation without assuming a specific theoretical distribution. Its basic idea is to generate samples from the observed data by sampling with replacement from the original data set. Using the empirical model and resampled data set in Experiment 721, the accuracy of the bootstrap regression model and ANN model can be improved.

[0288] Therefore, the research results and failure events and cases generated in the experiment 721 and the failure path test 722 can be used in the failure mechanism learning machine 730 to learn the event-driven state transition mechanism and determine the transfer function defined in the cluster machine.

[0289] For each identified risk threat, the risk modeling process 711 attempts to build a risk model for it. For cluster machines, the process first determines the possible failure paths with corresponding state transitions based on the situation from the failure event data server 703. Based on the event or cause and the input data set, the failure mechanism learning machine 730 also supports this process by providing accurate state and transition functions.

[0290] The constructed model should be verified by the verification process 723. The failure mechanism learning process 732 in the learning machine 730 can be used to compare multiple available models in the threat risk modeling 711 process.

[0291] The validation process 724 performs validity testing of the constructed model to verify whether the model reflects a variety of scenarios and a full range of application cases and conditions. The test can be multiple atomic simulations or coupled simulations. Prediction 712 is the system operation calculation output of risk modeling 711 and modeling machine 710. The result can be for one or more failure mode states (S f ∈S) is a set of failure probabilities or risk scores.

[0292] Figure 8 An exemplary embodiment of a framework for predicting the severity level of risk consequences is shown. The risk consequence calculation in the figure has two parts: system input 801 and system output 803. The system input 801 includes a set of consequence impact factors 802 and a set of failure mode states and their risk probabilities 816. The set of consequence impact factors 802 includes a receptor subset 821, a dispersion factor subset 822 and a product hazard subset 823.

[0293] The risk consequence is estimated based on the probability input of leakage 811, small leakage 812, large leakage 813, rupture 814 and breakdown 815 of the failure mode state and its risk probability set 816. The risk probability set is from Figure 7 The output data of prediction 712 is obtained.

[0294] The risk score output 803 by the system as the risk consequence assessment result is closely related to the risk severity level: the consequence of leakage 831 should be relatively minor, but the consequence of rupture 834 is much more serious than the consequence of small leakage 832, large leakage 833, and breakdown 835. All influencing factor sets 802 can also be pre-defined in this risk modeling system, and their index values ​​are modularized and adjusted according to actual conditions.

[0295] The receptor factor set 821 includes factors such as population density, environmental considerations, high value areas, etc. The dispersion factor set 822 includes weather, topography, surface flow resistance, volume of release, emergency response, and product characteristics.

[0296] Product Hazard Factor Set 823 includes both acute and chronic hazard characteristics. Some product hazard factor 823 is actually almost entirely acute, such as natural gas. However, both gas and liquid pipeline products should be evaluated based on their flammability (Nf), reactivity (Nr) and toxicity (Nh), which are acute hazards. Chronic hazard characteristics also include aquatic toxicity, mammalian toxicity, environmental persistence, flammability, corrosiveness and reactivity.

[0297] The multi-layer risk calculation modeling framework with failure mechanism learning machine is actually an open structure for all valid quantitative risk models. Both the simulation simulator and the failure mechanism learning machine can implement different risk models and experiment and learn them, and then verify and compare them in the modeling machine. In addition, the risk calculation modeling system is also open to new methods and algorithms for risk analysis and assessment.

[0298] In one or more aspects of the present invention, a processing circuit or unit includes at least one processing device, such as a microprocessor, a microcontroller, a digital signal processor, a microcomputer, a neural network, an AI processor, a quantum processor, a field programmable gate array, a programmable logic device, a state machine, a logic circuit, an analog circuit, a digital circuit and / or any device that manipulates signals (analog and / or digital) based on hard coding of circuits and / or operating instructions. The storage device used in this article is a non-temporary memory and can be an internal memory or an external memory, and the memory can be a single storage device or multiple storage devices. The memory can be a read-only memory, a random access memory and / or any non-temporary storage device that stores information. Typically, the processing circuit, unit or device is configured to perform one or more functions described herein in response to instructions stored in the storage device.

[0299] As may be used herein, the term "operable on" or "configurable on" or "configured to" indicates that an element includes one or more of circuits, instructions, modules, data, inputs, outputs, etc. to perform one or more of the described or necessary corresponding functions, and may also include inferred coupling to one or more other items to perform the described or necessary corresponding functions. As may also be used herein, the terms "coupled", "coupled to", "connected to" and / or "connected" or "interconnected" include direct connections or links between nodes / devices and / or indirect connections between nodes / devices through intermediate items (e.g., items including but not limited to components, elements, circuits, modules, nodes, devices, network elements, etc.). As may be further used herein, inferred connections (i.e., where one element is connected to another element by inference) include direct and indirect connections between two items in the same manner as "connected to".

[0300] As may be used herein, the terms "substantially" and "approximately" provide an industry-accepted tolerance of relativity between their corresponding terms and / or items. Such industry-accepted tolerances range from less than one percent to fifty percent and correspond to, but are not limited to, frequency, wavelength, component values, integrated circuit process variations, temperature variations, rise and fall times, and / or thermal noise. Correlations between items range from a few percentage points to orders of magnitude differences.

[0301] As used herein, the terms "comprises," "comprising," "having," "having," "including," "comprising," or any variations thereof, are intended to represent a non-exclusive inclusion such that a process, method, article, composition, or apparatus that includes a list of elements includes not only those elements recited but may include other elements not expressly listed or inherent to such process, method, article, composition, or apparatus. Except where not specifically recited, other combinations and / or modifications of the above-described structures, arrangements, applications, proportions, elements, materials, or components used in the practice of the invention may be varied or specially adapted to a particular environment, manufacturing specifications, design parameters, or other operating requirements without departing from the same general principles.

[0302] Furthermore, unless otherwise stated, reference to an element in the singular is not intended to mean "one and only one," but rather "one or more." Unless expressly stated otherwise, the term "some" refers to one or more. All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to one of ordinary skill in the art are expressly incorporated herein by reference and are intended to be covered by the claims. Furthermore, nothing disclosed herein is intended to be dedicated to the public, regardless of whether such disclosure is expressly recited in the claims.

[0303] The various features of the present disclosure described herein may be implemented in different systems and devices without departing from the present disclosure. It should be noted that the foregoing aspects of the present disclosure are merely examples and should not be construed as limiting the present disclosure. The description of the various aspects of the present disclosure is intended to be illustrative, rather than limiting, the scope of the claims. In this way, the present teachings may be easily applied to other types of devices, and many substitutions, modifications, and variations will be apparent to those skilled in the art.

[0304] In the foregoing description, certain representative aspects have been described with reference to specific examples. However, various modifications and changes may be made without departing from the scope of the claims. The description and drawings are illustrative rather than restrictive, and modifications are intended to be included within the scope of the claims. Therefore, the scope of the claims should be determined by the claims themselves and their legal equivalents, rather than by the examples described alone. For example, the components and / or elements recited in any device claim may be assembled or otherwise operably configured in various arrangements, and are therefore not limited to the specific configuration recited in the claims.

Claims

1. A system for pipeline operation and integrity management, the system comprising: Detection subsystem, including: a plurality of sensors located at a plurality of target locations in the pipeline, wherein the plurality of sensors generate field data measured at the target locations in the pipeline; an in-pipeline in-line inspection vehicle configured to traverse an enclosed portion of the pipeline between target locations, wherein the in-line inspection vehicle generates in-line data associated with the enclosed portion of the pipeline; and A companion device external to the pipeline, wherein the companion device is configured to track the position of the online inspection vehicle in the closed section of the pipeline and generate geographic data related to the position of the online inspection vehicle in the closed section of the pipeline, Data processing subsystem, including: at least one data store, including non-transitory memory, configured to store the online data and associated geographic data and to store the field data; and At least one data processing device configured to: Mapping online data from the online inspection vehicle to at least a portion of the pipeline being inspected using the generated geographic data associated with the online inspection vehicle; Determine the field data of the inspected portion of the pipeline; performing data mining on at least the field data and the online data of the inspected portion of the pipeline and generating at least input data of the inspected portion of the pipeline, Risk modeling subsystem, including: At least one modeling processing device is configured to implement risk model functionality based on a finite state machine, wherein the at least one modeling processing device processes input data of at least the inspected portion of the pipeline and generates a prediction of one of a plurality of failure mode states of the inspected portion of the pipeline.

2. The system according to claim 1, wherein: The data processing subsystem is also configured to: External data is generated using video external to the pipeline, wherein the video is obtained from at least one camera in the companion device, and wherein the external data is associated with GPS coordinates of the companion device.

3. The system according to claim 2, wherein: The data processing subsystem is further configured as follows: The online data, the field data and the external data are data processed to generate input data for at least the inspected portion of the pipeline for risk modeling.

4. The system according to claim 1, wherein: The online detection equipment includes a magnetic device configured to generate a magnetic field signal from inside the pipeline.

5. The system according to claim 4, wherein: The accompanying equipment outside the pipeline is set as: Detect the magnetic field signal generated by the online detection equipment inside the pipeline; Use magnetic field signals to determine the position of online detection equipment; generating geographic data associated with the location of the online inspection vehicle, wherein the geographic data includes GPS coordinates of the location of the online inspection device; and Associate geographic data with a timestamp.

6. The system according to claim 1, wherein: The data processing subsystem is also configured to: Pressure readings obtained from a plurality of pressure sensors in an online detection device are received and stored, wherein the plurality of pressure sensors are arranged on the online detection device and located at different positions around and near the circumference of the inner wall of the pipeline.

7. The system according to claim 6, wherein: The data processing subsystem is also configured to: determining a difference between two or more pressure readings obtained from the plurality of pressure sensors within approximately the same time frame; and When the difference is greater than a predetermined threshold, an indication of one of a plurality of failure mode conditions in the pipeline is generated.

8. The system according to claim 1, further comprising a management subsystem comprising: An operating subsystem is configured to initiate additional monitoring of the inspected portion of the pipeline by the in-line inspection vehicle in response to a prediction of one of the plurality of failure mode conditions of the inspected portion of the pipeline.

9. The system according to claim 1, wherein: The finite state machine based risk modeling functionality includes a cluster machine type processing model, and wherein the risk modeling subsystem is further configured to perform risk modeling of the pipeline by: Input data is processed using a cluster machine type processing model that defines multiple sets of transition functions between multiple initial states and multiple failure mode states, wherein the multiple sets of transition functions include one or more functions determined using one or more machine learning techniques, wherein known data inputs are used for known failure mode states.

10. The system according to claim 9, wherein: The cluster machine type processing model further defines a plurality of event clusters, wherein the plurality of event clusters at least include: an operation-related risk event cluster; a random risk event cluster; and a time-related risk event cluster.

11. The system according to claim 10, wherein: The cluster machine M is defined as: M= <X,S,Y,δ i ,δ e ,δ c ,λ,T,O,R,ta>, where X is the set of input values, S is the set of sequence states, Y is a set of output values, δ i is a set of internal transfer functions, δ e is a set of external transfer functions, δ c is a set of convergent transfer functions, T is the risk event cluster related to time, O is the risk event cluster related to operation, R is the random risk event cluster, λ is the output function set, and ta is the time advancement function.

12. A system for pipeline operation and integrity management, comprising: Detection subsystem, including: an in-line inspection vehicle configured to traverse an enclosed portion of the pipeline, wherein the in-line inspection vehicle acquires in-line data associated with the enclosed portion of the pipeline; and A companion device configured to move outside the pipeline and track the online inspection vehicle, wherein the companion device is configured to detect the position of the online inspection vehicle in the closed portion of the pipeline and generate geographic data related to the position of the online inspection vehicle; a data processing subsystem, comprising: at least one data store, including non-transitory memory, configured to store the online data and associated geographic data; and At least one data processing device, configured to: mapping the online data from the online inspection vehicle to the inspected portion of the pipeline using geographic data associated with the location of the online inspection vehicle to generate an online data set for the inspected portion of the pipeline; and Generate input data for the instrumented part of the pipeline using online datasets; The risk modeling subsystem comprises at least one modeling processing device, wherein the risk modeling subsystem is configured to: processing input data of at least the detected portion of the pipeline; and A prediction of one of a plurality of failure mode states of the inspected portion of the pipeline is generated.

13. The system according to claim 12, wherein: The data processing subsystem is also configured to: obtaining external data generated using video external to the pipeline, wherein the video is obtained from at least one camera in the companion device and the external data is associated with geographic data of the companion device; and Mapping external data to the inspected section of the pipeline using geographic data from the accompanying equipment; and An online dataset is generated for the instrumented portion of the pipeline using the external data and the online data mapped to the instrumented portion of the pipeline.

14. The system according to claim 13, wherein: The data processing subsystem is further configured as follows: Data processing is performed on the online data, the field data, and the external data to generate input data of at least the instrumented portion of the pipeline for risk modeling.

15. The system of claim 14, wherein: The data processing subsystem is also configured to: acquiring field data generated using a plurality of sensors at a target location in the pipeline, wherein the field data is measured at the target location in the pipeline; mapping any field data measured in the inspected portion of the pipeline using target locations of a plurality of sensors in the pipeline; and An online dataset is generated for the inspected portion of the pipeline using the field data, external data, and online data mapped to the inspected portion of the pipeline.

Images