Drilling report generation and review
The use of multimodal data and deep learning models for real-time drilling report generation and review addresses errors and inconsistencies, improving safety and reliability by enhancing monitoring and decision-making on drilling rigs.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- MATRIX JVCO LTD
- Filing Date
- 2024-12-20
- Publication Date
- 2026-06-25
AI Technical Summary
Current drilling report generation for drilling rigs is plagued by errors, inconsistencies, and delays, particularly in wellbore stability, equipment performance, and drilling fluid monitoring, leading to suboptimal decision-making and potential hazards.
A computer-implemented method utilizing multimodal data sources, including sensors, video, and deep learning models to generate and review drilling reports in real-time, reducing errors and inconsistencies, and enabling real-time monitoring of critical parameters.
Enhances operational reliability and safety by providing accurate, real-time monitoring and analysis of wellbore stability, equipment performance, and drilling fluids, reducing the likelihood of equipment failure and personnel injury.
Smart Images
Figure IB2024062997_25062026_PF_FP_ABST
Abstract
Description
[0001] December 20, 2024 MATRIX JVCO LTD trading as AIQ A1775O4WO CKA / HEP / Gsn
[0002] DRILLING REPORT GENERATION AND REVIEW
[0003] TECHNICAL FIELD
[0004] 5 This disclosure generally relates to computer-implemented methods, computer programs, and computing systems for generating and reviewing drilling reports for a drilling rig, especially an oil or gas drilling rig.
[0005] BACKGROUND 0
[0006] Drilling reports for drilling rigs are a critical component in documenting and assessing the operational performance of such rigs. As the drilling reports are used to improve operations, drilling reports have a significant impact on the operation performance of such rigs. 5
[0007] The accumulation of information from various data sources is an essential aspect of these reports, which are typically generated manually. However, this manual process can lead to a multitude of issues, including errors, delayed reporting, inconsistencies in description, or low precision of the drilling reports, as well as combinations thereof. 0
[0008] As a result, there exists a need for improved generation and review of drilling reports, particularly with regard to quality assurance and enhancing operational reliability. The current state of drilling report generation is often inadequate, resulting in suboptimal decision-making and potentially hazardous operations on drill rigs. 5
[0009] In particular, the generation of drilling reports related to wellbore stability and drilling fluids poses significant challenges. These reports must include data on critical parameters such as wellbore diameter and shape, drilling fluid properties (e.g., density and viscosity), circulation rate and pressure, temperature and pressure readings at0 various depths, and drill bit performance metrics (e.g., penetration rate and wear).
[0010] Improvements in this area can have a positive impact on drill rig safety by providing real-time monitoring of wellbore conditions and drilling fluid properties. This enables operators to adjust drilling parameters in response to changing wellbore conditions, thereby reducing the likelihood of equipment failure or personnel injury.
[0011] For instance, if a drilling report indicates that the wellbore is experiencing instability due to excessive pressure, an operator can adjust the circulation rate and pressure accordingly to prevent potential blowouts or wellbore collapse. Moreover, real-time monitoring of drill bit performance metrics allows operators to identify any issues with the drill bit before they lead to equipment failure or downtime.
[0012] Another critical aspect of drilling reports is equipment performance and maintenance. These reports must include data on parameters such as drill bit wear and performance metrics, mud pump performance (e.g., flow rate and pressure), motor performance (e.g., speed and torque), pressure and temperature readings at various depths, and maintenance records and schedules. Improvements in this area can have a positive impact on drill rig reliability by providing operators with real-time monitoring of equipment performance.
[0013] Therefore, improved drilling reports enable identification of potential issues before they lead to equipment failure or downtime, thereby reducing unexpected repairs or downtime. Additionally, effective scheduling and planning of maintenance activities are facilitated, which reduces the likelihood of equipment failures due to neglected maintenance tasks.
[0014] Furthermore, drilling reports related to drilling fluids and chemical additives pose significant challenges. These reports must include data on parameters such as drilling fluid properties (e.g., density and viscosity), chemical additives used in drilling fluids (e.g., lubricants and defoamers), dosage rates and concentrations of chemical additives, and laboratory test results for drilling fluid samples.
[0015] Improvements in this area can have a positive impact on drill rig safety by reducing the risk of adverse reactions or unexpected interactions between different chemical additives used in drilling fluids. Additionally, operators are enabled to make informed decisions about the use of chemical additives, thereby reducing the likelihood of equipment damage or personnel injury. In general, the advantage of improved drilling reports is the enhancement of operational reliability on drill rigs. This may in in particular be achieved through the provision of real-time monitoring and analysis of critical parameters related to wellbore stability, equipment performance, and drilling fluids and chemical additives. By facilitating better decision-making and more effective operations, improved drilling reports may contribute to improved safety, reliability, and overall performance of drill rigs.
[0016] In summary, the current state of drilling report generation poses significant challenges, including errors, delayed reporting, inconsistencies in description, or low precision of the drilling reports, as well as combinations thereof. Thus, a need exists for improved generation of drilling reports and for review of drilling reports, in particular for quality assurance and to improve operational reliability of the drilling rigs.
[0017] SUMMARY
[0018] The problem stated above is at least in part overcome by the following disclosure.
[0019] According to an embodiment, a computer-implemented method for providing a drilling report for a drilling rig is provided, comprising: obtaining data from a plurality of multimodal sources associated with the drilling rig, the plurality of multimodal sources selected from at least: sensor data, video data, acoustic data, text, voice.
[0020] The method may further comprise: receiving at least one deep learning model, multimodal processing of the data to generate input data for the at least one deep learning model, classifying the input data by the at least one deep learning model into at least one category selected from a plurality of categories, generating at least a part of the drilling report based on the at least one category.
[0021] The at least one category may comprise classifications of drilling activities for example related to the driller’s actions like different steps related to Coiled Tubing etc. The at least one category may be changed based on requirements, for example depending on the drilling rig and / or over time, for example as new technology is applied. The method may be adopted to changes in the at least one category, for example by re-training the at least one deep learning model.
[0022] This embodiment and the additional embodiments described below may enable the generation and review of high-quality drilling reports that may have reduced errors and / or inconsistencies or may even be free from errors and / or inconsistencies. These reports may provide real-time monitoring and analysis of critical parameters related to wellbore stability, equipment performance, and drilling fluids and chemical additives.
[0023] Thus, the embodiments may lead to a significant reduction in the likelihood of equipment failure or personnel injury due to inadequate monitoring or decisionmaking based on incomplete or inaccurate information. Furthermore, the operational reliability of drill rigs may be enhanced, which may lead to improved safety and / or reduced downtime.
[0024] The improved drilling reports by the computer-implemented methods disclosed herein may allow the enhancement of operational reliability on drill rigs through monitoring and analysis of critical parameters, for example related to wellbore stability, equipment performance, and drilling fluids and chemical additives. By facilitating better decisionmaking and more effective operations, improved drilling reports contribute to improved safety, reliability, and overall performance of drill rigs.
[0025] The computer-implemented method according to the embodiments may thus allow combination of multimodal data and improve accuracy of the drilling report. The drilling report may be used to ensure safety and / or productivity of the drilling rig.
[0026] Obtaining data may comprise obtaining real-time data. The multimodal processing of the data may be a real-time processing of the data, the classifying of the data may be a real-time classifying and / or the generating of the drilling report may be a real-time generating of the drilling report. The generating of the drilling report may be conducted without human input. In some embodiments, this may reduce the risk of wellbore collapse or blowouts. For example, by incorporating multimodal data such as video, acoustic, and / or sensor data, the method may provide real-time monitoring of wellbore conditions and drilling fluid properties, enabling automated adjusting or enabling operators to adjust drilling parameters in response to changing wellbore conditions and reducing the risk of equipment failure or personnel injury.
[0027] Using and processing of real-time data may in addition or alternatively increase the speed at which the drilling report is available. Drilling reports based on real-time data may thus further improve drilling rig reliability.
[0028] The generating of the drilling report may comprise: receiving human annotation input, extracting structured information, by a natural language processing module, from the human annotation input, generating a structured output based on the drilling report and the structured information, and presenting the structured output for human review.
[0029] This may reduce inconsistencies of the drilling report, for example if the human annotation input is received from different individuals.
[0030] The human annotation input may be associated with additional data associated with the drilling rig. The additional data may not be included in the data from the plurality of multimodal sources.
[0031] For example, the additional data maybe data not available for sensor input, for example if data sources are temporarily out of operation. In other examples, the additional data may be knowledge obtained by human interaction, for example the outcome of a safety briefing conducted by an individual.
[0032] The human annotation input may be associated with a subset of the data from the plurality of multimodal sources. The human annotation input may comprise voice data and / or text data.
[0033] This may simplify input for a human operator and contribute to the advantages mentioned herein. For both automated and semi-automated processes, the following embodiments may be applied.
[0034] The computer-implemented method may further comprise: segmenting the data into a plurality of temporal data segments, each temporal data segment associated with an operation time of the drilling rig. Classifying the data may comprise classifying the plurality of temporal data segments using the at least one deep learning model.
[0035] This may allow a time resolution aware classification of the data and may allow a time series information to be made available via the drilling report.
[0036] The at least one deep learning model may be trained on historical drilling data from at least one of the following: the drilling rig, a plurality of drilling rigs different from the drilling rig.
[0037] This may facilitate data set creation for training for the deep learning model.
[0038] The drilling report may be a daily drilling report, DDR.
[0039] This may have the advantage that a DDR is provided which seamlessly integrates with pre-existing reporting infrastructure.
[0040] The drilling report may comprise at least one of the following: descriptions of drilling activities, a start time of the drilling activity, an end time of the drilling activity, the at least one category.
[0041] This may allow for a detailed drilling report which increases reliability and / or operational safety of the drilling rig.
[0042] The computer-implemented method may further comprise: generating an alert based on the least one category to notify an operator of the drilling-rig of potential downtime risks and / or anomalies in drilling operations of the drilling rig. This may increase the operational reliability of the drilling rig. For example, this may enable operators to take proactive measures to mitigate potential downtime risks and anomalies in drilling operations, thereby reducing the likelihood of equipment failures or repairs. In addition or alternatively, this may improve operational reliability by providing real-time alerts and notifications to operators, allowing them to respond promptly to changes in drilling operations and prevent potential issues from escalating into major problems. Further in addition or alternatively, this may also enable more effective maintenance planning and scheduling, as operators can prioritize tasks based on the alerts and notifications received, reducing the likelihood of unexpected equipment failures or repairs and minimizing downtime.
[0043] The computer-implemented method may further comprise: providing a user interface to add additional annotations and / or perform corrections to the drilling report.
[0044] This may increase the ease of use for an individual editing or creating drilling reports. Thereby, the efficiency of reporting may be increased.
[0045] The plurality of multimodal sources may further comprise at least one of the following: external data sources, weather forecasts, geological surveys.
[0046] This may increase the relevance of the drilling report and / or increase the operational reliability or safety of the drilling rig.
[0047] The drilling report may comprise at least one activity code associated with a specific drilling activity, the activity code being selected from a predefined set of activity codes.
[0048] The predefined set of activity codes may include at least one of: a first code indicating a drilling activity, a second code indicating the specific task or operation being performed, and / or a third code indicating a duration.
[0049] For example, the specific tasks or operation being performed may comprise at least one of mud logging, drilling, and casing installation. Using the activity code and / or the predefined set of activity codes may ease standardization and analysis of the drilling report, thereby allowing comparisons with previous drilling reports and predictions regarding safety and operational reliability, for example using time series analysis.
[0050] The computer-implemented method may further comprise: providing an input field for reviewing and approving the drilling report.
[0051] This may increase the relevance of a drilling report. In some embodiments, providing means for reviewing and approving will increase the overall quality of the available dataset of drilling reports. This may in turn improve the quality of the available training data the at least one deep learning model.
[0052] In one embodiment, a computer-implemented method for reviewing a drilling report for a drilling rig is provided, the method comprising: receiving an existing drilling report, obtaining data from a plurality of multimodal sources associated with the existing drilling report, the plurality of multimodal sources selected from at least: sensor data, video data, acoustic data, text, voice;
[0053] The method may further comprise: receiving at least one deep learning model, multimodal processing of the data to generate input data for the at least one deep learning model, classifying the input data by the at least one deep learning model into at least one category selected from a plurality of categories, comparing the at least one category with at least one existing category in the existing drilling report, detecting discrepancies between the data and the existing drilling report based on the comparing, generating a finding report based on the discrepancies.
[0054] This may enable more accurate and efficient review of drilling reports, as the method leverages multimodal data processing and deep learning-based classification to identify discrepancies between actual operations and existing drilling reports.
[0055] This may lead to improved quality assurance and control of drilling operations, as operators may be notified of potential issues or anomalies based on the detected discrepancies potentially in real-time, allowing them to take corrective actions before they become major problems.
[0056] This may also reduce the likelihood of human error in creating or reviewing drilling reports, as the method consistently identifies discrepancies between data and existing reports, freeing up operators to focus on higher-level tasks.
[0057] This may ultimately result in improved overall safety and reliability of drilling operations, as discrepancies are identified and addressed promptly, reducing the risk of equipment failure or personnel injury.
[0058] In one embodiment, a computer program comprising instructions, which when executed by a computer, causing the computer to carry out the method of any one of the previous embodiments is provided.
[0059] In one embodiment, a computing system comprising means configured for carrying out the method of any of the previous embodiment is provided.
[0060] BRIEF DESCRIPTION OF THE DRAWINGS
[0061] Fig. 1 shows a drilling report according to an embodiment.
[0062] Fig. 2 shows a flowchart according to various embodiments.
[0063] Fig. 3 shows an automated drilling report generation according to an embodiment.
[0064] Fig. 4 shows a drilling report generation with human in the loop according to an embodiment.
[0065] Fig. 5 shows a computer-implemented method 700 for reviewing a drilling report for a drilling rig according to an embodiment.
[0066] Fig. 6 shows a computer-implemented method according to an embodiment. IO
[0067] Fig.7 shows a computer-implemented method according to a further embodiment.
[0068] DETAILED DESCRIPTION
[0069] This disclosure addresses a need for at least partially automated, accurate, and realtime generation of Drilling Reports which using a plurality of multimodal data sources. Drilling reports, DR, or daily drilling reports, DDR are routinely used for example within the oil and gas industry. Conventional DRs have a risk of human error, inconsistency, and delays due to their manual preparation by drilling supervisors and variation of the drilling report generation due to individual preferences. This may lead to deviations between DRs for example between different shifts on drilling rigs. Thus, a need exists for improved generation of drilling reports which at least partially overcomes these difficulties. Such improved DRs improve the reliability of operational data, thus contributing to safer, more efficient drilling operations and increasing reliability of the drilling rig.
[0070] Fig. 1 shows a drilling report according to an embodiment. As illustrated in Fig. 1, a drilling report too may comprise various multimodal data sources, for example time traces 102-120 of various sensor data and / or acoustic data.
[0071] While the time traces 102-120 are merely exemplary and may differ in data type, amplitude, frequency etc., such sensor data may comprise at least one of the following drilling data:
[0072] A hole depth, which may quantify the maximum vertical extent of the drilled wellbore, thereby potentially indicating the total potential reach of the drilling operation.
[0073] A bit depth, which may quantify a current vertical position of the drilling tool within the wellbore, which may provide real-time tracking of the drilling progress.
[0074] A hook load, which may quantify the weight suspended on the drilling block. The hook load may be crucial for understanding the mechanical stress and load characteristics during drilling. A block position which may tracks the vertical positioning of the block within the drilling rig. The block position may offer insights into the drilling rig's mechanical configuration and movement.
[0075] A flow rate which may quantify a volume of drilling mud being circulated into the wellbore per unit time. The flow rate maybe critical for maintaining hole stability and / or removing drill cuttings.
[0076] A stand pressure which may quantify the internal pressure experienced within the drilling column. The stand pressure may provide essential information about downhole conditions and potential challenges.
[0077] A rotation velocity which may quantify a rotational speed of the drilling tool. The rotation velocity may allow inferring drilling efficiency and penetration rate.
[0078] A torque which may quantify a rotational force generated by the interaction between the drilling tool and the surrounding rock formation. The torque may indicate drilling dynamics and potential mechanical resistances.
[0079] This list is not exhaustive, and further drilling related sensor data may be used.
[0080] Some activities that are harder to detect directly from the sensor data may require multimodal data, such as video data, acoustic data, text and / or voice, for example driller’s comments. Using such a plurality of multimodal sources may allow complementing the sensors data and draw a full picture of the drilling activities. Some of the multimodal sources may may at least in part be represented by time traces, like time traces 102-120 in Fig. 1, but other representations are possible. Some multimodal sources may not be represented as time traces.
[0081] The drilling report shown in Fig. 1 may be created by a method according to an embodiment, for example as illustrated with reference to Fig. 2 - Fig. 6 or any other embodiment disclosed herein.
[0082] Fig. 2 shows a flowchart for methods according to various embodiments. According to one embodiment explained with reference to Fig. 2 and Fig. 6, A computer-implemented method 6oo shown in Fig. 6 for providing a drilling report for a drilling rig is provided. The method 6oo comprises:
[0083] Obtaining, at 610, data from a plurality of multimodal sources. Exemplary multimodal source 210 are shown in Fig. 2. The multimodal sources 210 may be associated with the drilling rig. As shown in Fig. 1 and Fig. 2, the plurality of multimodal sources may be selected from at least: sensor data 102-120; 212, video data 214, acoustic data 216, text 130-140; 220, voice 220. In some embodiments, further data sources are used which may be of different formats and / or types, for example other modalities 218.
[0084] As non-limiting examples, plurality of multimodal sources may further comprise at least one of the following: external data sources, weather forecasts, geological surveys.
[0085] The method 600 further comprises: receiving, at 620, at least one deep learning model 234-
[0086] At 630, multimodal processing like multimodal processing 232 shown in Fig. 2 of the data to generate input data for the at least one deep learning model 234 may be conducted.
[0087] At 640, the method comprises classifying the input data by the at least one deep learning model 230 into at least one category selected from a plurality of categories. At step 650, the method comprises generating 250 at least a part of the drilling report 350, 480 based on the at least one category.
[0088] Obtaining data may comprises obtaining real-time data. „Real-time“ can mean close to the current time, for example with a delay of less than 10 seconds, for example less than 5 seconds, however, real-time may also mean near-real-time, for example in a subminute time frame, for example every 10 seconds, for example every 20 seconds, for example every 30 seconds, for example every 40 seconds, for example every 50 seconds, for example every minute, for example in intervals of minutes, for example every 5 minutes, for example every 10 minutes, for example every 20 minutes. The method disclose herein may significantly increase the data ingestion capabilities of a system, which may allow sub-minute time frame real-time data acquisition and processing.
[0089] In some embodiments, the multimodal processing 232 of the data is a real-time processing of the data, the classifying of the data may be a real-time classifying and / or the generating 250 of the drilling report may be a real-time generating 252 of the drilling report without human input.
[0090] In some embodiments, like shown in 252 in Fig. 2, a drilling report like the drilling report of Fig. 1 is generated automatically. An example is explained with reference to Fig. 3 below. In other embodiments, human annotation input may be received and included in the drilling report. An example is shown in 254 in Fig. 2, where a semiautomated drilling report is generated. This is detailed with reference to Fig. 4 below, but alternative implementations are possible as described herein. Further, in some embodiments, a computer-implemented method for reviewing a drilling report is provided. This may facilitate quality control 256 and is described with reference to Fig. 5 and Fig. 7 below.
[0091] Examples for automated drilling report generation
[0092] Fig. 3 shows an automated drilling report generation 300 according to an embodiment. In the example of Fig. 3, multimodal data, namely sensor data stream 312, video data 314, and acoustic data 316 is obtained, for example in a step similar to step 610 of Fig.
[0093] 6.
[0094] Fig. 3 further illustrates multimodal processing 332 of the data which results in classifying the input data by the at least one deep learning model. In Fig. 3, the result of the classifying is shown in reference numeral 332. Finally, a drilling report 350 is generated based on the at least one category. In the example of Fig. 3, the drilling report 350 is generated based on a plurality of categories shown in reference numeral 332. Semi-automated drilling report generation
[0095] Fig. 4 shows a drilling report generation with human in the loop according to an embodiment.
[0096] In such embodiments, the generating 250 of the drilling report may comprise: receiving human annotation input 440, 442 and extracting 430 structured information, for example by a natural language processing module 236, from the human annotation input. As shown in Fig. 4, the method may comprise generating a structured output 454 based on the drilling report and the structured information. Further, the method may comprise presenting 460 the structured output for human review.
[0097] The human annotation input 440, 442 may be associated with additional data 442 associated with the drilling rig. In some embodiments, the additional data 442 is not included in the data from the plurality of multimodal sources 210, 212-218; 410. This may allow the human to add additional information not available in the multimodal sources, for example information regarding a pre-job safety meeting in the text 480 as shown in Fig. 4. As also indicated by the text at reference numeral 480, the computer- implemented may provide an input field, for example a “review & edit” text field 480, for reviewing and approving the drilling report.
[0098] The human annotation input may be associated with a subset of the data from the plurality of multimodal sources, for example a temporal subset of data 102-120 is associated with annotation input 136 shown in Fig. 1 for this temporal time window indicated by box 136.
[0099] Further examples applicable to both automated and semi-automated drilling report generation
[0100] As explained with reference to Fig 1, but also in general, the method may comprise segmenting the data into a plurality of temporal data segments, each temporal data segment associated with an operation time of the drilling rig. For example, a first temporal data segment may be defined by box 132 in Fig. 1, a subsequent temporal data segment may be defined by box 134 in Fig. 1 etc. The classifying of the data may comprise classifying the plurality of temporal data segments using the at least one deep learning model 230. In some embodiments, the at least one deep learning model 230 may use a time series analysis. In other embodiments, the at least one deep learning model may use averaged data over the temporal data segments.
[0101] The at least one deep learning model 234 is trained on historical drilling data from at least one of the following: the drilling rig, a plurality of drilling rigs different from the drilling rig.
[0102] For examples, a plurality of drilling reports like drilling report too shown in Fig. 1 may be used to train the at least one deep learning model.
[0103] The method may further comprise: generating an alert based on the least one category to notify an operator of the drilling-rig of potential downtime risks and / or anomalies in drilling operations of the drilling rig.
[0104] As an illustrative example, based on the variations of sensors data and previously generated classifications and comments, the method may assess the activities and describe them when generating at least part of the drilling report. As shown in Fig. 1, the at least one deep learning modal may classify the data to show a slowdown of activity for approximately 30 minutes followed by a change in the direction from Pull Out Of Hole, POOH, 130 to Running In Hole, RIH, 130.
[0105] The classifying of the data may comprise classifying this sequence of the data with the corresponding activity codes, for example in Fig. 1 a first PJSM, Prejob Safety Meeting, followed by CT, Coiled Tubing. Based on the classifying, the respective description 130- 134 for these activities shown in Fig. 1 may be generated. The respective description may be part of the drilling report.
[0106] As disclosed herein, further multimodal sources associated with the drilling rig may be obtained, for example video data 214, acoustic data 216, and sensors data 102-120, 212. This may improve the accuracy of the generation of the drilling report. As a further illustrative example, based on weather data and on previous drilling reports 100 as shown in Fig. 1, the at least one learning model may classify weather data into categories “safe” and “risk”. However, further sub-classifications, for example a risk level, is possible. For example, from historical data like shut down of operation 140 shown in Fig. 1, historical classification by manual operation may be used to train the at least one deep learning model based on the data available. In some cases, the outcome of the decision is also known. For example, training data in which no shut in the well was carried out and subsequently weather damage occurred may be reclassified as “risk”. Thus, a set of training data including multimodal input data and classification data can be provided for training the at least one deep learning model. In the example of weather, also video data 214, acoustic data 216 and sensor data 102-120, 212 may be used to improve accuracy of the at least one deep learning model.
[0107] The drilling report may comprise at least one of the following: descriptions of drilling activities, a start time of the drilling activity, an end time of the drilling activity, the at least one category, for example one of the categories shown in Fig. 1 to Fig. 5.
[0108] The human annotation input may comprise voice data 440 and / or text data 442.
[0109] As shown in Fig. 4, the method may comprise providing a user interface 454, 460 to add additional annotations and / or perform corrections to the drilling report. In some embodiments, as shown in Fig. 2, the at least one deep learning model may comprise a model which provides a suggested auto completion 238. The auto completion 238 may then be displayed as a suggested text in the user interface 460. This may improve standardization of the drilling report data and improve quality and thus may increase operational reliability of the drilling rig.
[0110] Further examples regarding drilling report review
[0111] As discussed with reference to Fig. 2 above, the present disclosure may facilitate quality control 256. Turning to Fig. Fig. 5 and Fig. 7, Fig. 7 shows an embodiment of a method according to an embodiment and Fig. 5 shows a manual review process. The computer-implemented method 700 for reviewing a drilling report for a drilling rig may comprise:
[0112] Receiving, at 710, an existing drilling report, for example drilling report 510 shown in Fig. 5-
[0113] The method 700 may further comprise: obtaining 720 data from a plurality of multimodal sources 210 associated with the existing drilling report 510, the plurality of multimodal sources selected from at least: sensor data 212, video data 214, acoustic data 216, text 220, voice 220 etc. This may be similarly done as described herein.
[0114] The method 700 may further comprise: receiving 730 at least one deep learning model 234. The at least one deep learning model234 may also be identical to the at least one deep learning model used for the other embodiments described herein or it may be at least one deep learning model specifically trained for method 700.
[0115] The method 700 further comprises: multimodal processing, at 740, of the data to generate input data for the at least one deep learning model 234.
[0116] At 750, classifying of the input data by the at least one deep learning model 230 into at least one category selected from a plurality of categories may be performed.
[0117] At 760, comparing the at least one category with at least one existing category in the existing drilling report 510 may be carried out.
[0118] In 770, a detection of discrepancies between the data and the existing drilling report based on the comparing may be carried out. For example, Fig. 5 shows a summary of detected discrepancies 530. based on the discrepancies, the method 700 may then generate, at 780, a finding report, for example finding report 550 shown in Fig. 5.
[0119] The method 700 may allow a quality control, QC, which may allow for automated verification and / or evaluation of manually generated reports.
[0120] The method 700 may comprise cross-referencing descriptions, codes, and subcodes against multi-modal data to detect discrepancies. The method 700 may utilize the at least one deep learning model as described above. Thus, the method 700 may provide recommendations for activity descriptions and classifications, thereby improving operational reliability of the drilling rig.
[0121] The various embodiments described herein may thus increase the reliability of the drilling rig and / or increase the safety of the drilling rig. Further, the generated and / or quality-controlled drilling reports may result in increased efficiency, safety and / or reliability of the drilling rig. In particular in real-time embodiments, but not limited to real-time embodiments, the method according to various embodiments provide a diagnostic for an operational state of the drilling rig.
[0122] The algorithm according to the present invention may be implemented in terms of a computer program which may be executed on any suitable data processing device comprising means (e.g., a memoiy and one or more processors operatively coupled to the memory) being configured accordingly. The computer program may be stored as computer-executable instructions on a non-transitory computer-readable medium.
[0123] Embodiments of the present disclosure may be realized in any of various forms. For example, in some embodiments, the present invention may be realized as a computer- implemented method, a computer-readable memory medium, or a computer system.
[0124] In some embodiments, a non-transitory computer-readable memory medium may be configured so that it stores program instructions and / or data, where the program instructions, if executed by a computer system, cause the computer system to perform a method, e.g., any of the method embodiments described herein, or, any combination of the method embodiments de-scribed herein, or, any subset of any of the method embodiments described herein, or, any com-bination of such subsets.
[0125] In some embodiments, a computing device may be configured to include a processor (or a set of processors) and a memory medium, where the memory medium stores program instructions, where the processor is configured to read and execute the program instructions from the memory medium, where the program instructions are executable to implement any of the various method embodiments described herein (or, any combination of the method embodiments described herein, or, any subset of any of the method embodiments described herein, or, any combination of such subsets). The device may be realized in any of various forms.
[0126] Although specific embodiments have been described above, these embodiments are not intended to limit the scope of the present disclosure, even where only a single embodiment is described with respect to a particular feature. Examples of features provided in the disclosure are intended to be illustrative rather than restrictive unless stated otherwise. The above description is intended to cover such alternatives, modifications, and equivalents as would be apparent to a person skilled in the art having the benefit of this disclosure.
[0127] The scope of the present disclosure includes any feature or combination of features disclosed herein (either explicitly or implicitly), or any generalization thereof, whether or not it mitigates any or all of the problems addressed herein. In particular, with reference to the appended claims, features from dependent claims may be combined with those of the independent claims and features from respective independent claims may be combined in any appropriate manner and not merely in the specific combinations enumerated in the appended claims.
Claims
December 20, 2024 MATRIX JVCO LTD trading as AIQ A1775O4WO CKA / HEP / GsnCLAIMS1. A computer-implemented method (600) for providing a drilling report for a5 drilling rig, comprising: obtaining (610) data from a plurality of multimodal sources (210) associated with the drilling rig, the plurality of multimodal sources selected from at least: sensor data (102-120; 212, 312), video data (214, 314), 0 acoustic data (216, 316), text (130-140; 220), voice (220); receiving (620) at least one deep learning model (234), multimodal processing (630, 232) of the data to generate input data for the at5 least one deep learning model (234), classifying (640) the input data by the at least one deep learning model (230) into at least one category selected from a plurality of categories, generating (650, 250) at least a part of the drilling report (350, 480) based on the at least one category.
02. The computer-implemented method of claim 1, wherein: obtaining data comprises obtaining real-time data, the multimodal processing (232) of the data is a real-time processing of the data, 5 the classifying of the data is a real-time classifying and, the generating (250) of the drilling report is a real-time generating (252) of the drilling report without human input.
3. The computer-implemented method of claim 1, wherein the generating (250) of0 the drilling report comprises: receiving human annotation input (440, 442), extracting (430) structured information, by a natural language processing module (236), from the human annotation input,generating a structured output (454) based on the drilling report and the structured information, and presenting (460) the structured output for human review.
4. The computer-implemented method of claim 3, wherein the human annotation input (440, 442) is associated with additional data (442) associated with the drilling rig and wherein the additional data (442) is not included in the data from the plurality of multimodal sources (210, 212-218; 410).
5. The computer-implemented method of claim 3 or claim 4, wherein the human annotation input is associated with a subset of the data from the plurality of multimodal sources, and wherein the human annotation input comprises voice data (440) and / or text data (442).
6. The computer-implemented method of any of the previous claims, further comprising: segmenting the data into a plurality of temporal data segments, each temporal data segment associated with an operation time of the drilling rig, and classifying the data comprises classifying the plurality of temporal data segments using the at least one deep learning model (230).
7. The computer-implemented method of any of the previous claims, wherein the at least one deep learning model (234) is trained on historical drilling data from at least one of the following: the drilling rig, a plurality of drilling rigs different from the drilling rig.
8. The computer-implemented method of any of the previous claims, wherein the drilling report is a daily drilling report, DDR.
9. The computer-implemented method of any of the previous claims, wherein the drilling report comprises at least one of the following: descriptions of drilling activities,a start time of the drilling activity, an end time of the drilling activity, the at least one category. io. The computer-implemented method of any of the previous claims, further comprising: generating an alert based on the least one category to notify an operator of the drilling-rig of potential downtime risks and / or anomalies in drilling operations of the drilling rig. n. The computer-implemented method of any of the previous claims, further comprising: providing a user interface (454, 460) to add additional annotations and / or perform corrections to the drilling report.
12. The computer-implemented method of any of the previous claims, the plurality of multimodal sources further comprising at least one of the following: external data sources, weather forecasts, geological surveys.
13. The computer-implemented method of any of the previous claims, wherein the drilling report comprises at least one activity code associated with a specific drilling activity, the activity code being selected from a predefined set of activity codes.
14. The computer-implemented method of claim 13, wherein the predefined set of activity codes includes at least one of: a first code indicating a drilling activity, a second code indicating the specific task or operation being performed, and / or a third code indicating a duration.
15. The computer-implemented method of any of the previous claims, further comprising: providing an input field (480) for reviewing and approving the drilling report.
16. A computer-implemented method (700) for reviewing a drilling report for a drilling rig, comprising: receiving (710) an existing drilling report (510), obtaining (720) data from a plurality of multimodal sources (210) associated with the existing drilling report (510), the plurality of multimodal sources selected from at least: sensor data (212), video data (214), acoustic data (216), text (220), voice (220); receiving (730) at least one deep learning model (234), multimodal processing (740, 232) of the data to generate input data for the at least one deep learning model (234), classifying (750) the input data by the at least one deep learning model (230) into at least one category selected from a plurality of categories, comparing (760) the at least one category with at least one existing category in the existing drilling report (510), detecting (770) discrepancies (530) between the data and the existing drilling report based on the comparing, generating (780, 250) a finding report (550) based on the discrepancies.
17. A computer program comprising instructions, which when executed by a computer, causing the computer to carry out the method of any one of claims 1-15 or 16.
18. A computing system comprising means configured for carrying out the method of any one of claims 1-15 and / or claim 16.