WO2015106065A1 - Device and method for mitigating cycle-skipping in full waveform inversion - Google Patents
Device and method for mitigating cycle-skipping in full waveform inversion Download PDFInfo
- Publication number
- WO2015106065A1 WO2015106065A1 PCT/US2015/010768 US2015010768W WO2015106065A1 WO 2015106065 A1 WO2015106065 A1 WO 2015106065A1 US 2015010768 W US2015010768 W US 2015010768W WO 2015106065 A1 WO2015106065 A1 WO 2015106065A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- recorded
- data
- seismic data
- estimated
- model
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 103
- 230000000116 mitigating effect Effects 0.000 title description 7
- 238000001228 spectrum Methods 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000005315 distribution function Methods 0.000 claims description 6
- 230000000704 physical effect Effects 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 5
- 238000005314 correlation function Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 48
- 230000008569 process Effects 0.000 description 17
- 238000003908 quality control method Methods 0.000 description 17
- 238000013459 approach Methods 0.000 description 13
- 239000002245 particle Substances 0.000 description 11
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 4
- 230000003534 oscillatory effect Effects 0.000 description 4
- 230000007704 transition Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000001575 pathological effect Effects 0.000 description 2
- 229910052704 radon Inorganic materials 0.000 description 2
- SYUHGPGVQRZVTB-UHFFFAOYSA-N radon atom Chemical compound [Rn] SYUHGPGVQRZVTB-UHFFFAOYSA-N 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 101100087530 Caenorhabditis elegans rom-1 gene Proteins 0.000 description 1
- 101100305983 Mus musculus Rom1 gene Proteins 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003412 degenerative effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000005553 drilling Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000007429 general method Methods 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 150000002430 hydrocarbons Chemical class 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000003325 tomography Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
- G01V1/306—Analysis for determining physical properties of the subsurface, e.g. impedance, porosity or attenuation profiles
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/282—Application of seismic models, synthetic seismograms
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
- G01V1/303—Analysis for determining velocity profiles or travel times
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/34—Displaying seismic recordings or visualisation of seismic data or attributes
- G01V1/345—Visualisation of seismic data or attributes, e.g. in 3D cubes
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2200/00—Details of seismic or acoustic prospecting or detecting in general
- G01V2200/10—Miscellaneous details
- G01V2200/14—Quality control
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/61—Analysis by combining or comparing a seismic data set with other data
- G01V2210/614—Synthetically generated data
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
- G01V2210/622—Velocity, density or impedance
- G01V2210/6222—Velocity; travel time
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/66—Subsurface modeling
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/66—Subsurface modeling
- G01V2210/667—Determining confidence or uncertainty in parameters
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/67—Wave propagation modeling
Definitions
- Embodiments of the subject matter disclosed herein generally relate to methods and systems and, more particularly, to mechanisms and techniques for mitigating cycle-skipping in full waveform inversion (FWI).
- FWI full waveform inversion
- Seismic data acquisition and processing generate a profile (image) of the geophysical structure (subsurface) under the seafloor. While this profile does not provide an accurate location for oil and gas, it suggests, to those trained in the field, the presence or absence of oil and/or gas. Thus, providing a high-resolution image of the subsurface is an ongoing process for the exploration of natural resources, including, among others, oil and/or gas.
- a vessel 1 10 tows plural detectors 1 12, which are disposed along a cable 1 14. Cable 1 14 and its corresponding detectors 1 12 are sometimes referred to, by those skilled in the art, as a streamer 1 16.
- Vessel 1 10 may tow plural streamers 1 16 at the same time. Streamers may be disposed horizontally, i.e., lie at a constant depth zi relative to the ocean surface 1 18. Also, plural streamers 1 16 may form a constant angle (i.e., the streamers may be slanted) with respect to the ocean surface as disclosed in U.S. Patent No. 4,992,992, the entire content of which is incorporated herein by reference. In one embodiment, the streamers may have a curved profile as described, for example, in U.S. Patent No. 8,593,904, the entire content of which is incorporated herein by reference.
- vessel 1 10 may also tow a seismic source 120 configured to generate an acoustic wave 122a.
- Acoustic wave 122a propagates downward and penetrates the seafloor 124, eventually being reflected by a reflecting structure 126 (reflector R).
- Reflected acoustic wave 122b propagates upward and is detected by detector 1 12.
- Figure 1 shows only two paths 122a corresponding to the acoustic wave. Parts of reflected acoustic wave 122b (primary) are recorded by various detectors 1 12 (recorded signals are called traces), while parts of reflected wave 122c pass detectors 1 12 and arrive at the water surface 1 18. Since the interface between the water and air is well
- Wave 122d is normally referred to as a ghost wave because it is due to a spurious reflection. ghosts are also recorded by detector 1 12, but with a reverse polarity and a time lag relative to primary wave 122b if the detector is a hydrophone.
- the degenerative effect that ghost arrival has on seismic bandwidth and resolution is known. In essence, interference between primary and ghost arrivals causes notches, or gaps, in the frequency content recorded by detectors.
- Recorded traces may be used to determine the subsurface (i.e., earth structure below surface 124) and to determine the position and presence of reflectors 126, which is associated with the detection of oil and/or gas reservoirs.
- subsurface i.e., earth structure below surface 12
- reflectors 126 which is associated with the detection of oil and/or gas reservoirs.
- ghosts disturb the accuracy of the final image of the subsurface and, for at least this reason, various methods exist for removing the ghosts, i.e., deghosting, from the acquired seismic data.
- FWI is one established tool in the toolbox of velocity model building. A key practical issue in its successful application is avoiding convergence to a local minimum during the inversion process.
- the method includes receiving recorded seismic data recorded with seismic sensors over a subsurface of interest; selecting a model that describes the subsurface; calculating, based on the model and the recorded seismic data, estimated seismic data; and choosing a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data.
- the probabilistic measure includes at least one statistical function.
- the method includes receiving recorded seismic data d rec recorded with seismic sensors over a subsurface of interest;
- estimating seismic data d es t based on a starting model of the surveyed subsurface; calculating a travel-time difference (At) between the recorded data d rec and the estimated data d es t; converting the travel-time difference (At) to a phase difference; selecting seismic data with a phase difference between the recorded and estimated data of less than a half cycle; and feeding the selected seismic data to a full-wave inversion (FWI) model for calculating physical properties of the subsurface of interest.
- FWI full-wave inversion
- the computing device includes an interface configured to receive recorded seismic data recorded with seismic sensors over a subsurface of interest; and a processor connected to the interface.
- the processor is configured to select a model that describes the subsurface, calculate, based on the model and the recorded seismic data, estimated seismic data, and choose a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data.
- the probabilistic measure includes at least one statistical function.
- a computing device for determining a weight based on a data selection analysis.
- the computing device includes an interface configured to receive recorded seismic data d rec recorded with seismic sensors over a subsurface of interest; and a processor connected to the interface.
- the processor is configured to estimate seismic data d es t based on a starting model of the surveyed subsurface, calculate a travel-time difference (At) between the recorded data d rec and the estimated data d es t, convert the travel-time difference (At) to a phase difference, select seismic data with a phase difference between the recorded and estimated data of less than a half cycle, and feed the selected seismic data to a full-wave inversion (FWI) model for calculating physical properties of the subsurface of interest.
- FWI full-wave inversion
- Figure 1 is a schematic diagram of a conventional marine seismic data acquisition system having a horizontal streamer
- Figure 2 is a flowchart of an FWI method
- Figure 3 illustrates a recorded trace and an estimated trace
- Figure 4 illustrates the oscillatory nature of the cost function
- Figure 5A illustrates a cost function for a given time delay
- Figure 5B illustrates the probability distribution function of the measured time delay, also known as the probability density
- Figure 5C illustrates the probability distribution function of the measured frequency
- Figure 5D illustrates the probability of cycle-skipping versus time delay calculated based on a probabilistic measure according to an embodiment
- Figure 6 is a flowchart of a QC method for mitigating cycle-skipping based on a probabilistic measure
- Figure 7 is a flowchart of a QC method for mitigating cycle-skipping based on another probabilistic measure
- Figure 8 is a flowchart of a method that weights terms of a cost function based on a probabilistic measure
- Figure 9 is a flowchart of another method that weights terms of a cost function based on a data selection analysis
- Figure 10A illustrates recorded seismic data
- Figure 10B illustrates estimated seismic data
- Figure 10C illustrates a result of data selection that mitigates cycle-skipping
- Figure 1 1 is a schematic diagram of a computing device that can implement any of the methods discussed herein.
- Vp the seismic velocity of the compressional waves
- P-waves or primary waves the seismic velocity of the compressional waves
- Vp may be estimated in various ways.
- FWI is one known method for analyzing seismic data. FWI is able to produce models in a subsurface region of physical properties such as Vp that have high fidelity and are well-resolved spatially. FWI seeks to extract the properties of subsurface rocks from a given seismic dataset recorded at the surface or seabed.
- the FWI technique generates a two- or three-dimensional model to represent the surveyed subsurface and then attempts to control the properties and parameters of the Earth model to generate estimated seismic data that matches the recorded seismic data.
- the estimated seismic data may be calculated from the subsurface model based, for example, on the full two-way wave equation.
- FWI is an iterative process requiring a starting model.
- An accurate starting model for FWI may be provided by travel-time tomography.
- FWI can extract, for example, physical properties (Vp and Vs velocities, attenuation, density, anisotropy) of the surveyed subsurface.
- P-wave velocity Vp is a desired parameter on which the subsequent construction of the other parameters depends heavily upon.
- other parameters may be used in the following embodiments, either alone or in
- the model is used to calculate an estimate of the seismic dataset.
- the estimated seismic dataset is then compared to the recorded seismic dataset. Then, through use of a convergent numerical iteration, the parameters of the model are modified until the estimated seismic dataset matches the recorded seismic data to a sufficient degree of accuracy or until a desired degree of convergence is obtained.
- FWI typically operates on the principle of iteratively updating the starting model to minimize (or maximize) a cost function through repeated calculations.
- recorded seismic data is received.
- the recorded seismic data may be recorded on land or in a marine environment. It may be recorded with hydrophones, geophones, accelerometers, or any combination of these and other sensors.
- a model is chosen (e.g., selected from an existing library, calculated based on existing data about the surveyed subsurface, etc.) and seismic data is estimated based on the current model.
- a cost function is defined.
- the cost function may be a measure of the mismatch or similarity between the recorded seismic data and the estimated seismic data. If the cost function is selected to represent the mismatch between the recorded and estimated seismic data, it may compare traces from the two sets of data, for example, by subtracting one trace from another, i.e., one trace from each set of recorded and estimated seismic data.
- the cost function used in FWI may oscillate, rather than have a monotonic behavior. Because of this, it is necessary to have a sufficiently accurate starting model to achieve global minimum convergence.
- the cost function can be formulated in the frequency domain, the time domain or any other suitable domain.
- step 206 For Traditionally, localized gradient-based methods are used in step 206 for calculating the cost function. These methods iteratively update the existing model in a direction that derives from the cost function's direction of steepest descent. For this reason, after the cost function has been calculated in step 206, a given criterion (i.e., a predetermined condition) is checked in step 208. If the predetermined condition is not met, for example, how close the estimated data is to the recorded data, the process advances to step 210 in which the starting model is updated and a new data estimate is calculated. Then, the process returns to step 206 to recalculate the cost function.
- This FWI is a local iterative inversion scheme and the process makes in step 210 a series of step-wise improvements to the model, which successively reduces the cost function toward the predetermined condition.
- step 208 When the predetermined condition is met in step 208, the process advances to step 212, in which an accurate model has been obtained.
- the calculated model may then be used to calculate various parameters of the subsurface, generate an image of the subsurface, generate a velocity model, etc.
- FWI methodology for the cost function relies upon a gradient method to solve the inverse problem. This requires that the estimated data generated from the starting model match the observed travel times to within half a cycle. However, recorded seismic data is limited in its frequency bandwidth. This means that real seismic signals are oscillatory.
- Figure 3 shows a trace 300 corresponding to recorded seismic data and a trace 302 corresponding to estimated seismic data as a function of time (on the vertical axis). Trace 300 is received in step 200 and trace 302 is estimated in step 202.
- Figure 4 shows a cost function 400 that calculates a difference between traces 300 and 302 for a range of trial time shifts. As shown, the value of the cost function, or misfit, or objective function, is oscillatory as a function of this time shift. Thus, an inaccurate starting model may generate estimated data that is more than half a cycle in error with respect to the recorded data.
- cycle-skipped Such a situation is referred to as "cycle-skipped.”
- FWI will tend to modify the sub-surface model such that the estimated and recorded data are brought into alignment at the nearest cycle, and this will neither correspond to their correct alignment nor to the correct model. This misalignment to the nearest cycle will reduce the data misfit, and typical FWI schemes will become stranded in this position.
- the estimated seismic data needs to lie within half a wave cycle of the recorded data at the lowest useable frequency.
- the starting model needs to be sufficiently accurate to match the recorded seismic data to within half a wave cycle, otherwise FWI may mis-converge to a cycle-skipped local minimum.
- an approach based on a probabilistic interpretation of frequency and/or time-shift is now discussed with regard to a QC method.
- the QC method discussed herein detects traces that are likely to experience cycle-skipping. Synthetic or estimated traces are created using the starting velocity model, and the difference between the estimated and recorded data gives a cost function to be minimized by FWI. If the estimated traces have a delay time At compared to the recorded traces, and the time delay At is greater than (2 ⁇ f>) "1 , where ⁇ f> is the mean frequency of the data, then gradient methods will converge toward a local minimum and the updated velocity model will be biased.
- the mean frequency is calculated for a given trace. It is known that each trace includes all the frequencies in the seismic range of interest, e.g., between 1 and 300 Hz. Thus, the mean frequency for a given trace is expected to be similar to any other trace.
- the mean frequency is calculated, the power spectrum of a trace is represented versus the frequency, often the trace has been high-cut filtered to the lowest useable frequency, and this is used for calculating the mean frequency.
- a statistical estimator or operator or probabilistic measure may be used for the delay time.
- a probability distribution, or density, function PDF
- the PDF is characterized by the mean delay time ⁇ At> and a standard deviation OM-
- CDF cumulative distribution function
- CDF(At) PDF f ⁇ - ⁇ a ⁇ dT (1)
- ⁇ is a dummy time variable
- the vertical line inside equation (1 ) means that the PDF depends on OM and Of. An example in this sense is provided later. Note that in one embodiment, the CDF is calculated for each trace.
- Figure 5A illustrates the cost function used to infer the time delay between the recorded data and the estimated data.
- the global minimum represents the correct time delay while the other minima are cycle-skipped.
- the PDF for the time delay derived from Figure 5A is illustrated in Figure 5B.
- a threshold for Q can be chosen above which the risk of cycle-skipping is considered unacceptable and traces where Q exceeds this threshold are flagged so that the operator can decide how to handle those traces.
- the method includes a step 600 of receiving recorded seismic data (d rec ) recorded with seismic sensors over a subsurface of interest, a step 602 of selecting a model (m) that describes the subsurface, a step 604 of calculating, based on the model m and the recorded seismic data d re c, estimated seismic data (d es t), and a step 606 of choosing a probabilistic measure (e.g., the CDF or Q) that characterizes a relationship between the recorded seismic data d rec and the estimated seismic data d es t.
- the probabilistic measure includes at least one statistical function (PDF).
- the probabilistic measure is a probability that the corresponding trace will experience cycle-skipping.
- the QC procedure discussed above informs the operator which traces may create problems for the FWI methods, but does not remove or correct or update the traces to avoid these problems.
- the embodiments discussed above disclose how a probabilistic interpretation of measured time-shifts and frequency content can provide a robust and reliable QC of cycle-skipping in FWI, thus evaluating the suitability, or not, of a given starting model in 602.
- the probabilistic measure is capable of indicating data in a trace that can be used in the FWI while other data in the trace cannot.
- S(f) for a trace or for the entire recorded seismic data
- the mean frequency ⁇ f> is given by:
- time delay At is used to define the maximum frequency to be used for FWI purposes that avoids cycle-skipping. Note that each recorded trace includes all or most of the frequencies of interest, and the method discussed herein limits the frequency spectrum to be used for the FWI by calculating a new f max that prevents cycle-skipping. More specifically, the limit for f max is selected such that:
- the highest value of f ma x is chosen by integrating the amplitude spectrum S(f) so that the condition of equation (3) is fulfilled, which results in:
- the PDF includes uncertainties in the mean time delay (At) and the mean frequency (/), namely ⁇ ⁇ and a f .
- At mean time delay
- / mean frequency
- the predetermined threshold T may be selected, in one embodiment, such that any trace having its probability Q value larger than T is considered unacceptable for FWI, i.e., that trace is prone to cycle-skipping.
- predetermined threshold T is selected to be 0.05, or a 5% probability that the data is cycle-skipped. This is equivalent to a 2 ⁇ confidence level.
- the above-discussed method can be used to predict the maximum frequency to be used in the next update of FWI, whether this is the first FWI iteration, or a subsequent one.
- the value of f ma x varies along a survey. This means that the value of f ma x can change from trace to trace. Estimated shots from areas that are difficult to image are likely to show longer delay times At and, hence, lower values of f max .
- a likely strategy is then to create histograms of f max for the whole survey, and choose a value lower than that for a fraction of the survey. For example, a maximum frequency is chosen so that it is below the value of f ma x for 80% of the traces of the survey. It may not be practical to attempt a value low enough to meet the f ma x for 100% of the traces of the survey because this choice is likely to lead to an impractically low frequency.
- a method for identifying seismic traces prone to cycle-skipping in a full waveform inversion method includes a step 700 of receiving recorded seismic data recorded with seismic sensors over a subsurface of interest, a step 702 of selecting a model that describes the subsurface, a step 704 of calculating, based on the model and the recorded seismic data, estimated seismic data, a step 706 of choosing a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data, and a step 708 of calculating a maximum frequency f ma x based on the probabilistic measure, wherein frequencies below the maximum frequency are unlikely to have cycle-skipping.
- the wavelet has two characteristic timescales: (i) an outer envelope with a period inversely proportional to the bandwidth Af, and (ii) a generally higher-frequency oscillation with a period inversely proportional to the mean frequency (/).
- the probabilistic approach may be extended from a QC to an update in the cost function to weight down the cycle-skipped traces. This approach maximizes the amount of data used for FWI, since traces with a moderate probability of cycle- skipping are still used (albeit weighted down), rather than discarded.
- the method according to this embodiment potentially allows the removal of only the "pathological" traces at particular locations. For example, if in a subsurface region only the middle offsets have the probability Q of cycle-skipping larger than the predetermined threshold, the traditional cut-off strategy would remove all traces with offsets beyond the first pathological trace. This traditional approach correctly removes the middle offsets, but also removes the high offsets needlessly. Weighting down traces by their probability of cycle-skipping, as proposed in this embodiment, will remove the influence of the middle offsets while preserving the influence of the high offsets.
- the cost function is calculated as follows:
- n s , n r and n t are the number of sources, the number of receivers and the time samples in the dataset, respectively, and i, j, and k are the shot, receiver and time indices, respectively.
- weight w depends only on the shot index i and the receiver index j. Weight w may be calculated for every time sample in a seismic trace.
- Weight w may be implemented to include at least one probabilistic measure or function.
- Q is given by equations (5) and (6), although those skilled in the art would appreciate that other forms for Q may be used, or even other statistical functions instead of Q.
- traces that have a negligible probability of being cycle-skipped i.e., (3 * 0) are given a weight of substantially 1 .
- Traces with almost certainty of cycle-skipping i.e., Q « 1) are given zero weight, which is equivalent to entirely masking such areas.
- traces in the region with 0 ⁇ Q ⁇ 1 are given weights between 0 and 1 , and are therefore included in the calculation of the cost function and, hence, in the inversion. This means the data are used, but they would be weighted down. This extra data taken into account may be useful to improve the model on transition regions.
- FWI is often used to improve a model m (e.g., a velocity model) in difficult areas. Transition regions in between the areas where the model is good and bad are particularly important. For an FWI update, they can represent the region where the model transitions from not being cycle-skipped and, hence, safe to update, to being cycle-skipped and, hence, not updated meaningfully. Maximizing the use of data in such a region is therefore of particular interest because it will help the FWI update to gradually improve the model in such regions. Removing data with high probability of cycle-skipping avoids updating the model in the wrong direction on bad areas.
- m e.g., a velocity model
- selectively weighting the data can help update the transition regions, and the scheme discussed herein is designed to do exactly that, i.e., it enables the use of data in this region, while still allowing the cost function to be further influenced by regions where the current model is safe from cycle-skipping.
- a method for weighting seismic traces prone to cycle-skipping in a full waveform inversion includes a step 800 of receiving recorded seismic data recorded with seismic sensors over a subsurface of interest, a step 802 of selecting a model that describes the subsurface, a step 804 of calculating, based on the model and the recorded seismic data, estimated seismic data, a step 806 of choosing a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data, wherein the probabilistic measure includes at least one statistical function, a step 808 of calculating the probabilistic measure Q for a time delay At, between the recorded and estimated seismic data, and a step 810 of applying weights w to a cost function C, wherein the weights are related to the probabilistic measure Q, and the cost function C includes the recorded and estimated seismic data.
- step 900 receives in step 900 recorded seismic data d rec -
- the recorded seismic data may be recorded on land or water, with hydrophones, geophones, accelerometers, or a combination of them.
- the recorded seismic data may be single component (e.g., pressure) or multicomponent (e.g., velocity, acceleration, etc.).
- the method calculates in step 902 estimated seismic data d e st based on a starting model (e.g., velocity model) of the surveyed subsurface.
- step 904 calculates the travel-time difference (time delay At) between the recorded data d rec and estimated data d es t. This travel-time difference
- phase difference information can be converted to phase difference information.
- Data selection for mitigating cycle-skipping in FWI is then implemented by only using data with a phase difference between the recorded and estimated traces of less than a half cycle, namely ⁇ radians (or equivalently 180°). This process may be performed for each FWI iteration.
- the travel-time difference At is estimated using a windowed cross-correlation function.
- a windowed cross-correlation coefficient of two corresponding time sequences d rec (t) (recorded data) and d es t(t) (estimated data), with window size N and time-lags I is calculated in step 906 using the expression:
- Ci — ⁇ ⁇ -, (11) where (d rec ⁇ d est ) is the inner product of the two sequences d rec and d es t, (d rec ) and a drec are the average and standard deviation within the window, respectively, and d est l is the sequence with lag of I.
- the cross-correlation coefficient c is maximized when the two sequences best match in both time shift and event registration.
- step 908 By calculating q(x) near the first event (along a given trace) of each offset x and picking the first maximum peak l(x) in step 908, it is possible to estimate a guide line of the travel-time difference between the recorded data and estimated data. This repeated process of forming followed by smoothing a guide of the travel- time differences allows the influence of noisy traces to be down-weighted.
- the guide itself is a set of time-lags that "guides" the next iteration of the cross-correlation process.
- a linear smoothing filter may then be applied in step 910 to the picked guide over the different offsets, as follows: where:
- the weighting factor w(x 0 , x) is determined by the distance between the center point x 0 and reference point x with exponential decay factor d, and the reference cross-correlation coefficient q(x) with probability factor p.
- a higher cross-correlation coefficient suggests a more reliable estimation and is spread to its neighbors for smoothing.
- adaptive moving window to perform automatic data selection, progressively compensating time difference in each window, calculating a similarity between recorded data and estimated data in each adaptive window, with continuous constrain of neighboring windows, and labeling and muting bad data based on time difference and similarity criteria.
- Other weighting schemes in this linear smoothing filter can be used to achieve the same goal as equation (13).
- a second picking from the given trace is performed in step 912, near the smoothed guide line, to provide a more accurate estimation.
- This procedure may be iterative; however, in some tests performed based on this procedure, one iteration was usually sufficient.
- the process which may be implemented on a computing device as discussed later, performs automatic data selection in step 914, where only data with a phase difference of less than half cycle is selected to reduce the effect of cycle-skipping.
- the conversion of the travel-time difference into a phase difference may happen, for example, at step 914, just before the data selection criterion is applied.
- This data selection procedure includes two steps: discard bad traces and mute bad data (inner mute).
- the travel-time difference estimated above is used to determine bad traces via the phase difference limitation.
- an adaptive moving window is applied to calculate the similarity between the recorded data and estimated data, where the time difference down the trace is progressively
- c(t, k) is the cross-correlation coefficient of equation (1 1 ) calculated with lag k in the window at d rec (x,t) and d es t(x, t+l(x,t)), with window size of At.
- compensations are subject to continuous constraint to minimize cycle-skipping between neighboring windows, i.e., I(x,t)-I(x, t-At) ⁇ diff, where diff is the maximum allowed travel-time difference between the neighboring windows.
- discontinuities reflect unmatched or cycle-skipped events, and are used to improve the data selection.
- the time shifts l(x,t) are converted in step 914 to phase by multiplying them by frequency. Data with a phase difference of more than half cycle are muted. Then, in step 916, the FWI is applied to the selected data.
- Figure 10C shows an example of a data selection result. Even when the travel-time difference is very small, the data mismatch between the two windows of recorded data (see Figure 10A) and estimated data (see Figure 10B) can still occur due to missing events or crossing of refraction and reflection energy.
- the matching quality is also partially determined by the maximum coefficient c(t, k) of each window, and those below a specified threshold are labeled as bad and muted.
- hydrophone and/or particle velocity data collected from receivers are processed according to one or more of the embodiments discussed above.
- various known processing steps e.g., migration, FWI, denoising, deghosting, etc.
- an oil and gas reservoir interpreter determines whether oil and/or gas is present in the surveyed subsurface and advises an oil and gas company where to drill a well.
- the embodiments discussed above improve this technological process of detecting oil and/or gas reservoirs by mitigating cycle- skipping, which results in a more accurate subsurface image and hence reduced cost associated with the drilling process.
- receivers generate a marine streamer dataset that is achieved in a narrow, wide or multi-azimuth coil shooting or any configuration towed with constant or variable depth (e.g., slant streamer, BroadSeis profile, over- under streamers), and the seismic data is generated with an air gun, marine vibrator, or other source element.
- Source elements may be fired according to any known scheme, e.g., continuously, simultaneously, flip-flop, etc.
- Receivers may also be used in ocean bottom survey (nodes, cables, or other with air gun, marine vibrator or other source), land dataset (dynamite, vibrator or other source), or a combination of two or more dataset types.
- the data may have been calibrated before applying the processes discussed herein, or calibration scalars may be included in the
- Water velocity terms may be constant or allowed to vary with depth. Variation with depth can be of use for datasets where there is a water velocity gradient.
- the methods may be used for one-sided or split- spread acquisition.
- Equations described herein may be solved in the time domain or a spectral domain (e.g., frequency, Laplace, z-transform, etc.), wavelet domain (e.g., curvelet or other).
- Model m may be found through any inversion method, e.g., conjugate gradients, LU decomposition, Cholesky factorization, etc.
- Model m may be derived to represent all traces in the input shot, or may work on a subset of data from the input shot, for example, spatial windows of a given number of channels.
- Sparseness weights may be used in the inversion to improve results, for example, where there is poor signal-to-noise ratio or to overcome aliasing; e.g., iteratively reweighted least squares beginning with low frequencies and working up to higher frequencies.
- Other model domains may be used, for example, frequency- wavenumber (FK), parabolic Radon, hyperbolic Radon, etc.
- FK frequency- wavenumber
- parabolic Radon parabolic Radon
- hyperbolic Radon etc.
- any fixed datum model domain may be defined as long as it can be reverse-transformed, redatumed and reghosted for both hydrophone and particle velocity sensor data.
- an iterative approach similar to the anti-leakage tau-p transform can be used, which also exhibits sparseness properties. No matter how the model is formed, it needs to simultaneously reproduce the hydrophone and particle velocity measurements through application of an operator, e.g., L.
- a generalized weighting scheme can be implemented to vary weighting between any component (hydrophone or particle motion) depending on the receiver, time and frequency, and/or wavenumber. This weighting refers to how well the model represents the data.
- accelerometer data is integrated to calculate particle velocity measurements. Instead of integrating accelerometer data before wave-field separation, it is also possible to build a differentiation operator into the inversion scheme.
- wave-field extrapolation may be used, which may include Kirchhoff, beam, wave-equation (one-way or two-way, RTM, etc.).
- FIG. 1 1 The above-discussed procedures and methods may be implemented in a computing device as illustrated in Figure 1 1 .
- Hardware, firmware, software or a combination thereof may be used to perform the various steps and operations described herein.
- Computing device 1 100 of Figure 1 1 is an exemplary computing structure that may be used in connection with such a system.
- Exemplary computing device 1 100 suitable for performing the activities described in the exemplary embodiments may include a server 1 101 .
- a server 1 101 may include a central processor (CPU) 1 102 coupled to a random access memory (RAM) 1 104 and to a read-only memory (ROM) 1 106.
- ROM 1 106 may also be other types of storage media to store programs, such as programmable ROM (PROM), erasable PROM (EPROM), etc.
- Processor 1 102 may communicate with other internal and external components through input/output (I/O) circuitry 1 108 and bussing 1 1 10 to provide control signals and the like.
- I/O input/output
- Bussing 1 1 10 to provide control signals and the like.
- Processor 1 102 carries out a variety of functions as are known in the art, as dictated by software and/or firmware instructions.
- Server 1 101 may also include one or more data storage devices, including hard drives 1 1 12, CD-ROM drives 1 1 14 and other hardware capable of reading and/or storing information, such as DVD, etc.
- software for carrying out the above-discussed steps may be stored and distributed on a CD- ROM or DVD 1 1 16, a USB storage device 1 1 18 or other form of media capable of portably storing information. These storage media may be inserted into, and read by, devices such as CD-ROM drive 1 1 14, disk drive 1 1 12, etc.
- Server 1 101 may be coupled to a display 1 120, which may be any type of known display or presentation screen, such as LCD, plasma display, cathode ray tube (CRT), etc.
- a user input interface 1 122 is provided, including one or more user interface mechanisms such as a mouse, keyboard, microphone, touchpad, touch screen, voice-recognition system, etc.
- Server 1 101 may be coupled to other devices, such as sources, detectors, etc.
- the server may be part of a larger network configuration as in a global area network (GAN) such as the Internet 1 128, which allows ultimate connection to various landline and/or mobile computing devices.
- GAN global area network
- the disclosed exemplary embodiments provide a computing device, software instructions and a method for seismic data processing. It should be understood that this description is not intended to limit the invention. On the contrary, the exemplary embodiments are intended to cover alternatives,
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
Computing device, computer instructions and method for identifying seismic traces prone to cycle-skipping in a full waveform inversion method. The method includes receiving (600) recorded seismic data recorded with seismic sensors over a subsurface of interest; selecting (602) a model that describes the subsurface; calculating (604), based on the model and the recorded seismic data, estimated seismic data; and choosing (606) a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data. The probabilistic measure includes at least one statistical function.
Description
DEVICE AND METHOD FOR MITIGATING CYCLE-SKIPPING IN FULL
WAVEFORM INVERSION
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] The present application is related to and claims the benefit of priority of U.S. Provisional Application 61 /925,709 filed January 10, 2014, titled, "CSKIP: A probabilistic QC for Cycle-Skipping in Full Waveform Inversion," authored by A.
Sansigre and A. Ratcliffe; U.S. Provisional Application 61/927,194, filed January 14, 2014, titled, "Effective Cycle Skipping Reduction Through Adaptive Data Selection for Full Waveform Inversion," authored by H. Bi and T. Lin; U.S. Provisional
Application 61/945,843 filed February 28, 2014, titled, "Determining the Maximum Frequency to be used in a time-domain FWI update of a velocity model," authored by A. Sansigre and A. Ratcliffe; and U.S. Provisional Application 62/021 ,027 filed July 4, 2014, titled, "A probabilistic weighting scheme for the Full Waveform Inversion cost function," authored by D'afonseca et. al., the entire contents of which are
incorporated herein by reference.
BACKGROUND
TECHNICAL FIELD
[0002] Embodiments of the subject matter disclosed herein generally relate to methods and systems and, more particularly, to mechanisms and techniques for mitigating cycle-skipping in full waveform inversion (FWI).
DISCUSSION OF THE BACKGROUND
[0003] Seismic data acquisition and processing generate a profile (image) of the geophysical structure (subsurface) under the seafloor. While this profile does not provide an accurate location for oil and gas, it suggests, to those trained in the field, the presence or absence of oil and/or gas. Thus, providing a high-resolution image of the subsurface is an ongoing process for the exploration of natural resources, including, among others, oil and/or gas.
[0004] During a seismic gathering process, as shown for instance in the marine case in Figure 1 , a vessel 1 10 tows plural detectors 1 12, which are disposed along a cable 1 14. Cable 1 14 and its corresponding detectors 1 12 are sometimes referred to, by those skilled in the art, as a streamer 1 16. Vessel 1 10 may tow plural streamers 1 16 at the same time. Streamers may be disposed horizontally, i.e., lie at a constant depth zi relative to the ocean surface 1 18. Also, plural streamers 1 16 may form a constant angle (i.e., the streamers may be slanted) with respect to the ocean surface as disclosed in U.S. Patent No. 4,992,992, the entire content of which is incorporated herein by reference. In one embodiment, the streamers may have a curved profile as described, for example, in U.S. Patent No. 8,593,904, the entire content of which is incorporated herein by reference.
[0005] Still with reference to Figure 1 , vessel 1 10 may also tow a seismic source 120 configured to generate an acoustic wave 122a. Acoustic wave 122a propagates downward and penetrates the seafloor 124, eventually being reflected by
a reflecting structure 126 (reflector R). Reflected acoustic wave 122b propagates upward and is detected by detector 1 12. For simplicity, Figure 1 shows only two paths 122a corresponding to the acoustic wave. Parts of reflected acoustic wave 122b (primary) are recorded by various detectors 1 12 (recorded signals are called traces), while parts of reflected wave 122c pass detectors 1 12 and arrive at the water surface 1 18. Since the interface between the water and air is well
approximated as a quasi-perfect reflector (i.e., the water surface acts as a mirror for acoustic waves), reflected wave 122c is reflected back toward detector 1 12 as shown by wave 122d in Figure 1 . Wave 122d is normally referred to as a ghost wave because it is due to a spurious reflection. Ghosts are also recorded by detector 1 12, but with a reverse polarity and a time lag relative to primary wave 122b if the detector is a hydrophone. The degenerative effect that ghost arrival has on seismic bandwidth and resolution is known. In essence, interference between primary and ghost arrivals causes notches, or gaps, in the frequency content recorded by detectors.
[0006] Recorded traces may be used to determine the subsurface (i.e., earth structure below surface 124) and to determine the position and presence of reflectors 126, which is associated with the detection of oil and/or gas reservoirs. However, ghosts disturb the accuracy of the final image of the subsurface and, for at least this reason, various methods exist for removing the ghosts, i.e., deghosting, from the acquired seismic data.
[0007] FWI is one established tool in the toolbox of velocity model building. A key practical issue in its successful application is avoiding convergence to a local minimum during the inversion process. A large part of this local minimum problem is caused by the oscillatory nature of the seismic data and manifests itself as cycle- skipping on the data, with the model converging to the wrong event in the shot record. There is academic and industry activity in this area, although the standard way of addressing this problem is the hierarchical multi-scale approach proposed by Bunks et al. (1995, "Multiscale seismic wave-form inversion," Geophysics, 60, 1 ,457- 1 ,473). In this method, successive inversions are performed with an increasing high- frequency cut-off value, and the assumption is that the starting point in this process, namely the data at the lowest available frequency, is not cycle-skipped.
[0008] However, existing methods are not always efficient or, if they are, they are expensive. Thus, there is a need to develop new FWI methods that mitigate cycle-skipping, or to develop QC (quality control) methods that can indicate data prone to cycle-skipping prior to running the FWI algorithms.
SUMMARY
[0009] According to an embodiment, there is a method for identifying seismic traces prone to cycle-skipping in a full waveform inversion method. The method includes receiving recorded seismic data recorded with seismic sensors over a subsurface of interest; selecting a model that describes the subsurface; calculating,
based on the model and the recorded seismic data, estimated seismic data; and choosing a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data. The probabilistic measure includes at least one statistical function.
[0010] According to another embodiment, there is a method for determining a weight based on a data selection analysis. The method includes receiving recorded seismic data drec recorded with seismic sensors over a subsurface of interest;
estimating seismic data dest based on a starting model of the surveyed subsurface; calculating a travel-time difference (At) between the recorded data drec and the estimated data dest; converting the travel-time difference (At) to a phase difference; selecting seismic data with a phase difference between the recorded and estimated data of less than a half cycle; and feeding the selected seismic data to a full-wave inversion (FWI) model for calculating physical properties of the subsurface of interest.
[001 1] According to yet another embodiment, there is a computing device for identifying seismic traces prone to cycle-skipping in a full waveform inversion method. The computing device includes an interface configured to receive recorded seismic data recorded with seismic sensors over a subsurface of interest; and a processor connected to the interface. The processor is configured to select a model that describes the subsurface, calculate, based on the model and the recorded seismic data, estimated seismic data, and choose a probabilistic measure that
characterizes a relationship between the recorded seismic data and the estimated seismic data. The probabilistic measure includes at least one statistical function.
[0012] According to still another embodiment, there is a computing device for determining a weight based on a data selection analysis. The computing device includes an interface configured to receive recorded seismic data drec recorded with seismic sensors over a subsurface of interest; and a processor connected to the interface. The processor is configured to estimate seismic data dest based on a starting model of the surveyed subsurface, calculate a travel-time difference (At) between the recorded data drec and the estimated data dest, convert the travel-time difference (At) to a phase difference, select seismic data with a phase difference between the recorded and estimated data of less than a half cycle, and feed the selected seismic data to a full-wave inversion (FWI) model for calculating physical properties of the subsurface of interest.
BRIEF DESCRIPTION OF THE DRAWINGS
[0013] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate one or more embodiments and, together with the description, explain these embodiments. In the drawings:
[0014] Figure 1 is a schematic diagram of a conventional marine seismic data acquisition system having a horizontal streamer;
[0015] Figure 2 is a flowchart of an FWI method;
[0016] Figure 3 illustrates a recorded trace and an estimated trace;
[0017] Figure 4 illustrates the oscillatory nature of the cost function;
[0018] Figure 5A illustrates a cost function for a given time delay, Figure 5B illustrates the probability distribution function of the measured time delay, also known as the probability density, Figure 5C illustrates the probability distribution function of the measured frequency, and Figure 5D illustrates the probability of cycle-skipping versus time delay calculated based on a probabilistic measure according to an embodiment;
[0019] Figure 6 is a flowchart of a QC method for mitigating cycle-skipping based on a probabilistic measure;
[0020] Figure 7 is a flowchart of a QC method for mitigating cycle-skipping based on another probabilistic measure;
[0021] Figure 8 is a flowchart of a method that weights terms of a cost function based on a probabilistic measure;
[0022] Figure 9 is a flowchart of another method that weights terms of a cost function based on a data selection analysis;
[0023] Figure 10A illustrates recorded seismic data, Figure 10B illustrates estimated seismic data and Figure 10C illustrates a result of data selection that mitigates cycle-skipping; and
[0024] Figure 1 1 is a schematic diagram of a computing device that can implement any of the methods discussed herein.
DETAILED DESCRIPTION
[0025] The following description of the exemplary embodiments refers to the accompanying drawings. The same reference numbers in different drawings identify the same or similar elements. The following detailed description does not limit the invention. Instead, the scope of the invention is defined by the appended claims.
[0026] Reference throughout the specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with an embodiment is included in at least one embodiment of the subject matter disclosed. Thus, the appearance of the phrases "in one embodiment" or "in an embodiment" in various places throughout the specification is not necessarily referring to the same embodiment. Further, the particular features, structures or characteristics may be combined in any suitable manner in one or more embodiments.
[0027] As noted in the Background section, land or marine seismic surveys can be used to extract rock properties and construct reflectivity images of the subsurface. Such surveys can provide an accurate image of the subsurface structure of the portion of the earth being surveyed. The subsurface structure may be associated with mineral resources and/or hydrocarbons reservoirs. To model and image the subsurface, one needs to know at least the seismic velocity Vp of the compressional waves, also known as P-waves or primary waves. In a portion of the volume of the Earth, Vp may be estimated in various ways.
[0028] FWI is one known method for analyzing seismic data. FWI is able to produce models in a subsurface region of physical properties such as Vp that have high fidelity and are well-resolved spatially. FWI seeks to extract the properties of subsurface rocks from a given seismic dataset recorded at the surface or seabed.
[0029] Essentially, the FWI technique generates a two- or three-dimensional model to represent the surveyed subsurface and then attempts to control the properties and parameters of the Earth model to generate estimated seismic data that matches the recorded seismic data.
[0030] The estimated seismic data may be calculated from the subsurface model based, for example, on the full two-way wave equation. FWI is an iterative process requiring a starting model. An accurate starting model for FWI may be provided by travel-time tomography. FWI can extract, for example, physical properties (Vp and Vs velocities, attenuation, density, anisotropy) of the surveyed subsurface. P-wave velocity Vp is a desired parameter on which the subsequent construction of the other parameters depends heavily upon. However, other parameters may be used in the following embodiments, either alone or in
combination. The nature and number of parameters used in a model associated with a subsurface depends on the application.
[0031] The model is used to calculate an estimate of the seismic dataset. The estimated seismic dataset is then compared to the recorded seismic dataset. Then, through use of a convergent numerical iteration, the parameters of the model are
modified until the estimated seismic dataset matches the recorded seismic data to a sufficient degree of accuracy or until a desired degree of convergence is obtained.
[0032] A general method of updating the model with the FWI method is now described with regard to Figure 2. FWI typically operates on the principle of iteratively updating the starting model to minimize (or maximize) a cost function through repeated calculations. In step 200, recorded seismic data is received. The recorded seismic data may be recorded on land or in a marine environment. It may be recorded with hydrophones, geophones, accelerometers, or any combination of these and other sensors. In step 202, a model is chosen (e.g., selected from an existing library, calculated based on existing data about the surveyed subsurface, etc.) and seismic data is estimated based on the current model. In step 204, a cost function is defined. The cost function may be a measure of the mismatch or similarity between the recorded seismic data and the estimated seismic data. If the cost function is selected to represent the mismatch between the recorded and estimated seismic data, it may compare traces from the two sets of data, for example, by subtracting one trace from another, i.e., one trace from each set of recorded and estimated seismic data.
[0033] Due to the non-linearity in the relationship between the model and the data, the cost function used in FWI may oscillate, rather than have a monotonic behavior. Because of this, it is necessary to have a sufficiently accurate starting
model to achieve global minimum convergence. The cost function can be formulated in the frequency domain, the time domain or any other suitable domain.
[0034] Traditionally, localized gradient-based methods are used in step 206 for calculating the cost function. These methods iteratively update the existing model in a direction that derives from the cost function's direction of steepest descent. For this reason, after the cost function has been calculated in step 206, a given criterion (i.e., a predetermined condition) is checked in step 208. If the predetermined condition is not met, for example, how close the estimated data is to the recorded data, the process advances to step 210 in which the starting model is updated and a new data estimate is calculated. Then, the process returns to step 206 to recalculate the cost function. This FWI is a local iterative inversion scheme and the process makes in step 210 a series of step-wise improvements to the model, which successively reduces the cost function toward the predetermined condition.
[0035] Note that iteration steps 210 and 206 are necessary because the problem to be solved is non-linear, and the inverse problem has been linearized in particular stages. Implicitly, traditional inverse methods invoke the Born
approximation, which assumes that the perturbation to a wave-field produced by changing a model is linearly related to the change in the model.
[0036] When the predetermined condition is met in step 208, the process advances to step 212, in which an accurate model has been obtained. The
calculated model may then be used to calculate various parameters of the subsurface, generate an image of the subsurface, generate a velocity model, etc.
[0037] The above method, if converging to the global minimum, will lead to a model which is a correct representation of the surveyed subsurface. However, there are some difficulties associated with obtaining the desired convergence.
[0038] As discussed above, FWI methodology for the cost function relies upon a gradient method to solve the inverse problem. This requires that the estimated data generated from the starting model match the observed travel times to within half a cycle. However, recorded seismic data is limited in its frequency bandwidth. This means that real seismic signals are oscillatory.
[0039] This problem is illustrated based on Figures 3 and 4, which correspond to Figures 4 and 5 of U.K. Patent Application GB 2509223. Figure 3 shows a trace 300 corresponding to recorded seismic data and a trace 302 corresponding to estimated seismic data as a function of time (on the vertical axis). Trace 300 is received in step 200 and trace 302 is estimated in step 202. Figure 4 shows a cost function 400 that calculates a difference between traces 300 and 302 for a range of trial time shifts. As shown, the value of the cost function, or misfit, or objective function, is oscillatory as a function of this time shift. Thus, an inaccurate starting model may generate estimated data that is more than half a cycle in error with respect to the recorded data. Such a situation is referred to as "cycle-skipped." When this occurs, because the methodology seeks only a local minimum, FWI will
tend to modify the sub-surface model such that the estimated and recorded data are brought into alignment at the nearest cycle, and this will neither correspond to their correct alignment nor to the correct model. This misalignment to the nearest cycle will reduce the data misfit, and typical FWI schemes will become stranded in this position.
[0040] Consequently, as illustrated above, for successful convergence, the estimated seismic data needs to lie within half a wave cycle of the recorded data at the lowest useable frequency. In other words, the starting model needs to be sufficiently accurate to match the recorded seismic data to within half a wave cycle, otherwise FWI may mis-converge to a cycle-skipped local minimum.
[0041] As discussed above, the origin of cycle-skipping lies in the industrial FWI implementations that use a gradient-based algorithm that can only converge to the minimum in the model domain, that is, "local" to the starting point. While more sophisticated inversion algorithms exist that do not suffer the drawbacks of gradient- based techniques, they are much more expensive, and their application is beyond current practical implementations of FWI in the seismic industry.
[0042] Determining whether a given velocity model is good enough becomes an exercise in checking whether every trace in the shot record is not cycle-skipped. While this is sufficient to ensure convergence to the global minimum, given the nonlinear and iterative nature of FWI, having every trace non-cycle-skipped is not a necessary condition. It was shown that a basic phase difference between recorded
and estimated data can be used to highlight areas of cycle-skipping (Shah et al., 2012, "Quality assured full-waveform inversion: Ensuring starting model adequacy," 82nd Annual International Meeting, SEG, Expanded Abstracts, 1-5). Such a quality control (QC) method requires some degree of interpretation to extract the necessary information.
[0043] According to an embodiment, an approach based on a probabilistic interpretation of frequency and/or time-shift is now discussed with regard to a QC method. Note that the QC method discussed herein detects traces that are likely to experience cycle-skipping. Synthetic or estimated traces are created using the starting velocity model, and the difference between the estimated and recorded data gives a cost function to be minimized by FWI. If the estimated traces have a delay time At compared to the recorded traces, and the time delay At is greater than (2<f>)"1, where <f> is the mean frequency of the data, then gradient methods will converge toward a local minimum and the updated velocity model will be biased. Thus, a QC approach is therefore to check that the time delay is smaller than (2<f>)" 1. Note that according to an embodiment, the mean frequency is calculated for a given trace. It is known that each trace includes all the frequencies in the seismic range of interest, e.g., between 1 and 300 Hz. Thus, the mean frequency for a given trace is expected to be similar to any other trace. When the mean frequency is calculated, the power spectrum of a trace is represented versus the frequency, often
the trace has been high-cut filtered to the lowest useable frequency, and this is used for calculating the mean frequency.
[0044] Both the time delay At and the mean frequency <f> will suffer from uncertainties, and these should be taken into account when quantifying the risk of cycle-skipping in the inversion. According to an embodiment, a statistical estimator or operator or probabilistic measure may be used for the delay time. For example, a probability distribution, or density, function (PDF) is used as the estimator, and the PDF is characterized by the mean delay time <At> and a standard deviation OM- Interpreting the amplitude spectrum (i.e., distribution of power amplitude versus frequency) of the estimated data as the PDF of the frequency, it is possible to derive a mean frequency <f> and associated uncertainty Of. Then, it is possible to derive the cumulative distribution function (CDF) for the probability of cycle-skipping which, according to an implementation, is given by:
At
CDF(At) = PDF f^ - ^ a^ dT (1) where τ is a dummy time variable, and the vertical line inside equation (1 ) means that the PDF depends on OM and Of. An example in this sense is provided later. Note that in one embodiment, the CDF is calculated for each trace.
[0045] The above-discussed concepts are illustrated in Figures 5A-D. Figure 5A illustrates the cost function used to infer the time delay between the recorded data and the estimated data. The global minimum represents the correct time delay
while the other minima are cycle-skipped. The PDF for the time delay derived from Figure 5A is illustrated in Figure 5B. The amplitude spectrum of the seismic data interpreted as the PDF of the frequency is illustrated in Figure 5C, and Figure 5D illustrates the probability of cycle-skipping, Q (= 1 - CDF), as a function of the time delay. Note that Q does not suffer from wrap-around, i.e., for large values of the time delay, it continues to have a value of 1 . The time delay corresponding to Q = 0.5 is shown in Figure 5D as an indication of the cycle skipping criterion if there were no uncertainties present. A threshold for Q can be chosen above which the risk of cycle-skipping is considered unacceptable and traces where Q exceeds this threshold are flagged so that the operator can decide how to handle those traces.
[0046] A method that calculates a probabilistic QC measure for cycle-skipping in full waveform inversion is now discussed with regard to Figure 6. The method includes a step 600 of receiving recorded seismic data (drec) recorded with seismic sensors over a subsurface of interest, a step 602 of selecting a model (m) that describes the subsurface, a step 604 of calculating, based on the model m and the recorded seismic data drec, estimated seismic data (dest), and a step 606 of choosing a probabilistic measure (e.g., the CDF or Q) that characterizes a relationship between the recorded seismic data drec and the estimated seismic data dest. The probabilistic measure includes at least one statistical function (PDF). As noted above, the probabilistic measure (Q in this embodiment) is a probability that the corresponding trace will experience cycle-skipping.
[0047] Note that the QC procedure discussed above informs the operator which traces may create problems for the FWI methods, but does not remove or correct or update the traces to avoid these problems. Thus, the embodiments discussed above disclose how a probabilistic interpretation of measured time-shifts and frequency content can provide a robust and reliable QC of cycle-skipping in FWI, thus evaluating the suitability, or not, of a given starting model in 602.
[0048] According to another embodiment, it is possible to determine a probabilistic measure that relies on the measured time delay At between the recorded data drec and estimated data dest and indicates what is the maximum frequency that can be used without compromising the inversion. This means that the probabilistic measure is capable of indicating data in a trace that can be used in the FWI while other data in the trace cannot. In this regard, given an amplitude spectrum S(f) for a trace or for the entire recorded seismic data, the mean frequency <f> is given by:
Γ fmax
(f) = J S(f)df, (2)
fmin
where fmin and fmax are the minimum and maximum frequencies of interest. If the time delay is fulfilling At≥ (2(/))_1, then gradient-based algorithms will converge toward a local minimum rather than the global one and, hence, the updated velocity model will be cycle-skipped.
[0049] Thus, according to an embodiment, time delay At is used to define the maximum frequency to be used for FWI purposes that avoids cycle-skipping. Note that each recorded trace includes all or most of the frequencies of interest, and the method discussed herein limits the frequency spectrum to be used for the FWI by calculating a new fmax that prevents cycle-skipping. More specifically, the limit for fmax is selected such that:
[0050] In one implementation, the highest value of fmax is chosen by integrating the amplitude spectrum S(f) so that the condition of equation (3) is fulfilled, which results in:
J rfmax
S(J)df < (2ty (4)
However, note that there are other methods of deriving the characteristic frequency, such as a predictive error filter, etc.
[0051] According to another embodiment, a probabilistic method that implements equation (3) is now exemplified. Those skilled in the art would know that other probabilistic functions may be used to implement equation (3). The probability Q of the time delay At exceeding a predetermined threshold is given by probabilistic measure Q: PDF(r)dT, (5)
where τ is a dummy time variable. The PDF includes uncertainties in the mean time delay (At) and the mean frequency (/), namely σΔί and af . One possible form of the PDF may be:
Those skilled in the art would appreciate that other forms for the PDF may be used, or even other statistical functions instead of the PDF.
[0052] The predetermined threshold T may be selected, in one embodiment, such that any trace having its probability Q value larger than T is considered unacceptable for FWI, i.e., that trace is prone to cycle-skipping. Thus, the time tT at which Q(tT) = T may be calculated. In one embodiment, the value of the
predetermined threshold T is selected to be 0.05, or a 5% probability that the data is cycle-skipped. This is equivalent to a 2σ confidence level. Using equation (4), the maximum frequency fmax may expressed as:
(f) < {2tTy (7)
[0053] The above-discussed method can be used to predict the maximum frequency to be used in the next update of FWI, whether this is the first FWI iteration, or a subsequent one. In practice, it is likely that the value of fmax varies along a survey. This means that the value of fmax can change from trace to trace. Estimated shots from areas that are difficult to image are likely to show longer delay times At and, hence, lower values of fmax. A likely strategy is then to create histograms of fmax
for the whole survey, and choose a value lower than that for a fraction of the survey. For example, a maximum frequency is chosen so that it is below the value of fmax for 80% of the traces of the survey. It may not be practical to attempt a value low enough to meet the fmax for 100% of the traces of the survey because this choice is likely to lead to an impractically low frequency.
[0054] According to an embodiment, a method for identifying seismic traces prone to cycle-skipping in a full waveform inversion method includes a step 700 of receiving recorded seismic data recorded with seismic sensors over a subsurface of interest, a step 702 of selecting a model that describes the subsurface, a step 704 of calculating, based on the model and the recorded seismic data, estimated seismic data, a step 706 of choosing a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data, and a step 708 of calculating a maximum frequency fmax based on the probabilistic measure, wherein frequencies below the maximum frequency are unlikely to have cycle-skipping.
[0055] While the various embodiments discussed above have targeted a QC approach to the cycle-skipping problem, the next embodiments address the cost function of the FWI procedure and present probabilistic approaches to be
implemented in the cost function to mitigate the cycle-skipping problem.
[0056] As previously discussed, FWI implementations using gradient-based algorithms converge to the global minimum only if the starting point is close enough. More sophisticated inversion algorithms do exist, but they are more expensive and
their application is currently beyond industrial implementations of FWI. The criterion for convergence towards the global minimum is known in the art, namely, that the estimated and recorded data agree within half a period. If not, then FWI will converge to a wrong model (a local minimum) even though the cost function will decrease. Determining whether a given velocity model is good enough becomes an exercise in checking whether the estimated traces are not cycle-skipped with respect to the recorded ones. A basic phase difference between recorded and estimated data can be used to highlight areas of cycle-skipping. Such a QC requires some degree of interpretation to extract the necessary information, which is undesired. Thus, one or more of the embodiments here present a different approach that is based on a probabilistic interpretation of frequency and time-shift.
[0057] As discussed above, estimated traces are created using the starting velocity model, and their difference with the recorded data gives a cost function to be minimized by FWI. Because the amplitude spectrum of the traces has a limited bandwidth (the seismic frequency spectrum of interest is substantially between 1 and 300 Hz) and non-zero mean frequency, in the time domain, the wavelet has two characteristic timescales: (i) an outer envelope with a period inversely proportional to the bandwidth Af, and (ii) a generally higher-frequency oscillation with a period inversely proportional to the mean frequency (/). If the estimated traces have a delay time At compared to the recorded data that is greater than (2(/))_1, then gradient methods will converge toward a local minimum, and the updated velocity
model will be biased. Embodiments discussed above used probabilistic QC methods (see Figures 6 and 7) for identifying cycle-skipping in the traces.
[0058] The probabilistic approach may be extended from a QC to an update in the cost function to weight down the cycle-skipped traces. This approach maximizes the amount of data used for FWI, since traces with a moderate probability of cycle- skipping are still used (albeit weighted down), rather than discarded.
[0059] In addition, compared to the strategy of cutting off data when a threshold probability of cycle-skipping is reached, the method according to this embodiment potentially allows the removal of only the "pathological" traces at particular locations. For example, if in a subsurface region only the middle offsets have the probability Q of cycle-skipping larger than the predetermined threshold, the traditional cut-off strategy would remove all traces with offsets beyond the first pathological trace. This traditional approach correctly removes the middle offsets, but also removes the high offsets needlessly. Weighting down traces by their probability of cycle-skipping, as proposed in this embodiment, will remove the influence of the middle offsets while preserving the influence of the high offsets.
[0060] One particular way of weighting the cost function is now discussed. Those skilled in the art would readily understand that other ways of weighting the cost function may be derived from this particular implementation. Thus, this implementation is intended to be an exemplary approach for a better understanding
of the general concept of weighting down the unusable data from the FWI point of view.
[0061] In one implementation, the cost function is calculated as follows:
ns nr nt
C (m) =— ^ ^ ^ I dest i j k — drec i j k I (8)
i j k
where m is the model parameter (e.g., velocity), ns, nr and nt are the number of sources, the number of receivers and the time samples in the dataset, respectively, and i, j, and k are the shot, receiver and time indices, respectively.
[0062] To introduce the probabilistic approach, a weight w that depends on indices i, j and k is introduced in equation (8) as follows:
ns nr nt
C(m) = ^ ^ w(i,j, k) \dest iijik - dreCii ik \2 (9)
i j k
In one embodiment, weight w depends only on the shot index i and the receiver index j. Weight w may be calculated for every time sample in a seismic trace.
[0063] Weight w may be implemented to include at least one probabilistic measure or function. For example, weight w(i,j) may be implemented to be a probability that a given trace is not cycle-skipped. Such an example is described as follows: w(i,j) = 1 - Q(At =
Of ), (10) where Q is the probability that a trace is cycle-skipped. In one application, Q is given by equations (5) and (6), although those skilled in the art would appreciate that other
forms for Q may be used, or even other statistical functions instead of Q.
Irrespective of the configuration of Q, the probability Q that the trace is cycle-skipped is given by calculating Q at \ At\ = (2(/))~\ as also expressed by equation (10).
[0064] According to this embodiment, traces that have a negligible probability of being cycle-skipped (i.e., (3 * 0) are given a weight of substantially 1 . Traces with almost certainty of cycle-skipping (i.e., Q « 1) are given zero weight, which is equivalent to entirely masking such areas. However, traces in the region with 0 < Q < 1 are given weights between 0 and 1 , and are therefore included in the calculation of the cost function and, hence, in the inversion. This means the data are used, but they would be weighted down. This extra data taken into account may be useful to improve the model on transition regions.
[0065] In other words, FWI is often used to improve a model m (e.g., a velocity model) in difficult areas. Transition regions in between the areas where the model is good and bad are particularly important. For an FWI update, they can represent the region where the model transitions from not being cycle-skipped and, hence, safe to update, to being cycle-skipped and, hence, not updated meaningfully. Maximizing the use of data in such a region is therefore of particular interest because it will help the FWI update to gradually improve the model in such regions. Removing data with high probability of cycle-skipping avoids updating the model in the wrong direction on bad areas. Thus, selectively weighting the data can help update the transition regions, and the scheme discussed herein is designed to do exactly that, i.e., it
enables the use of data in this region, while still allowing the cost function to be further influenced by regions where the current model is safe from cycle-skipping.
[0066] According to an embodiment, a method for weighting seismic traces prone to cycle-skipping in a full waveform inversion includes a step 800 of receiving recorded seismic data recorded with seismic sensors over a subsurface of interest, a step 802 of selecting a model that describes the subsurface, a step 804 of calculating, based on the model and the recorded seismic data, estimated seismic data, a step 806 of choosing a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data, wherein the probabilistic measure includes at least one statistical function, a step 808 of calculating the probabilistic measure Q for a time delay At, between the recorded and estimated seismic data, and a step 810 of applying weights w to a cost function C, wherein the weights are related to the probabilistic measure Q, and the cost function C includes the recorded and estimated seismic data.
[0067] According to another embodiment illustrated in Figure 9, there is another method for applying a weighting to the cost function based on a data selection analysis. This method receives in step 900 recorded seismic data drec- The recorded seismic data may be recorded on land or water, with hydrophones, geophones, accelerometers, or a combination of them. In other words, the recorded seismic data may be single component (e.g., pressure) or multicomponent (e.g., velocity, acceleration, etc.). The method calculates in step 902 estimated seismic
data dest based on a starting model (e.g., velocity model) of the surveyed subsurface. In step 904, the method calculates the travel-time difference (time delay At) between the recorded data drec and estimated data dest. This travel-time difference
information can be converted to phase difference information. For example, if a single frequency f is considered, the phase difference φ (measured in radians) is given by φ = 2π■ At■ f, where the time delay At is measured in seconds and the frequency f is measured in Hz. Data selection for mitigating cycle-skipping in FWI is then implemented by only using data with a phase difference between the recorded and estimated traces of less than a half cycle, namely π radians (or equivalently 180°). This process may be performed for each FWI iteration.
[0068] According to this embodiment, the travel-time difference At is estimated using a windowed cross-correlation function. A windowed cross-correlation coefficient of two corresponding time sequences drec(t) (recorded data) and dest(t) (estimated data), with window size N and time-lags I is calculated in step 906 using the expression:
1 (drec (drec) \ dest, {dest)j
Ci =— ■ ■ -, (11) where (drec \ dest) is the inner product of the two sequences drec and dest, (drec) and adrec are the average and standard deviation within the window, respectively, and dest l is the sequence with lag of I. Those skilled in the art would appreciate that the inner product can be implemented in various ways. The cross-correlation coefficient
c; is maximized when the two sequences best match in both time shift and event registration. By calculating q(x) near the first event (along a given trace) of each offset x and picking the first maximum peak l(x) in step 908, it is possible to estimate a guide line of the travel-time difference between the recorded data and estimated data. This repeated process of forming followed by smoothing a guide of the travel- time differences allows the influence of noisy traces to be down-weighted. The guide itself is a set of time-lags that "guides" the next iteration of the cross-correlation process.
[0069] A linear smoothing filter may then be applied in step 910 to the picked guide over the different offsets, as follows:
where:
In the above equations, the weighting factor w(x0, x) is determined by the distance between the center point x0 and reference point x with exponential decay factor d, and the reference cross-correlation coefficient q(x) with probability factor p. A higher cross-correlation coefficient suggests a more reliable estimation and is spread to its neighbors for smoothing. For example, it is possible to apply adaptive moving window to perform automatic data selection, progressively compensating time
difference in each window, calculating a similarity between recorded data and estimated data in each adaptive window, with continuous constrain of neighboring windows, and labeling and muting bad data based on time difference and similarity criteria. Other weighting schemes in this linear smoothing filter can be used to achieve the same goal as equation (13).
[0070] A second picking from the given trace is performed in step 912, near the smoothed guide line, to provide a more accurate estimation. This procedure may be iterative; however, in some tests performed based on this procedure, one iteration was usually sufficient. Next, the process, which may be implemented on a computing device as discussed later, performs automatic data selection in step 914, where only data with a phase difference of less than half cycle is selected to reduce the effect of cycle-skipping. In one embodiment, the conversion of the travel-time difference into a phase difference may happen, for example, at step 914, just before the data selection criterion is applied.
[0071] This data selection procedure includes two steps: discard bad traces and mute bad data (inner mute).
[0072] The travel-time difference estimated above is used to determine bad traces via the phase difference limitation. For the selected traces, an adaptive moving window is applied to calculate the similarity between the recorded data and estimated data, where the time difference down the trace is progressively
compensated by the previous windows based on equation:
l(x, t) = l(x, t - At) + org max{c(t, k)}, (14)
k
where c(t, k) is the cross-correlation coefficient of equation (1 1 ) calculated with lag k in the window at drec(x,t) and dest(x, t+l(x,t)), with window size of At. The
compensations are subject to continuous constraint to minimize cycle-skipping between neighboring windows, i.e., I(x,t)-I(x, t-At) < diff, where diff is the maximum allowed travel-time difference between the neighboring windows. Those
discontinuities reflect unmatched or cycle-skipped events, and are used to improve the data selection. The time shifts l(x,t) are converted in step 914 to phase by multiplying them by frequency. Data with a phase difference of more than half cycle are muted. Then, in step 916, the FWI is applied to the selected data.
[0073] Figure 10C shows an example of a data selection result. Even when the travel-time difference is very small, the data mismatch between the two windows of recorded data (see Figure 10A) and estimated data (see Figure 10B) can still occur due to missing events or crossing of refraction and reflection energy. The matching quality is also partially determined by the maximum coefficient c(t, k) of each window, and those below a specified threshold are labeled as bad and muted.
[0074] According to an embodiment, hydrophone and/or particle velocity data collected from receivers (located on streamers, ocean bottom cables, autonomous vehicles, etc.) are processed according to one or more of the embodiments discussed above. Note that in the field of oil and gas, there are well-established processes for taking seismic raw data and applying various known processing steps,
e.g., migration, FWI, denoising, deghosting, etc., for generating an image of the surveyed subsurface. Based on this image, which can be represented in print, on a screen, in a tabular way, etc., an oil and gas reservoir interpreter determines whether oil and/or gas is present in the surveyed subsurface and advises an oil and gas company where to drill a well. The embodiments discussed above improve this technological process of detecting oil and/or gas reservoirs by mitigating cycle- skipping, which results in a more accurate subsurface image and hence reduced cost associated with the drilling process.
[0075] In one embodiment, receivers generate a marine streamer dataset that is achieved in a narrow, wide or multi-azimuth coil shooting or any configuration towed with constant or variable depth (e.g., slant streamer, BroadSeis profile, over- under streamers), and the seismic data is generated with an air gun, marine vibrator, or other source element. Source elements may be fired according to any known scheme, e.g., continuously, simultaneously, flip-flop, etc. Receivers may also be used in ocean bottom survey (nodes, cables, or other with air gun, marine vibrator or other source), land dataset (dynamite, vibrator or other source), or a combination of two or more dataset types. The data may have been calibrated before applying the processes discussed herein, or calibration scalars may be included in the
formulations noted in the embodiments. Water velocity terms may be constant or allowed to vary with depth. Variation with depth can be of use for datasets where
there is a water velocity gradient. The methods may be used for one-sided or split- spread acquisition.
[0076] Equations described herein may be solved in the time domain or a spectral domain (e.g., frequency, Laplace, z-transform, etc.), wavelet domain (e.g., curvelet or other). Model m may be found through any inversion method, e.g., conjugate gradients, LU decomposition, Cholesky factorization, etc. Model m may be derived to represent all traces in the input shot, or may work on a subset of data from the input shot, for example, spatial windows of a given number of channels. Sparseness weights may be used in the inversion to improve results, for example, where there is poor signal-to-noise ratio or to overcome aliasing; e.g., iteratively reweighted least squares beginning with low frequencies and working up to higher frequencies. Other model domains may be used, for example, frequency- wavenumber (FK), parabolic Radon, hyperbolic Radon, etc. In fact, any fixed datum model domain may be defined as long as it can be reverse-transformed, redatumed and reghosted for both hydrophone and particle velocity sensor data. Alternatively, an iterative approach similar to the anti-leakage tau-p transform can be used, which also exhibits sparseness properties. No matter how the model is formed, it needs to simultaneously reproduce the hydrophone and particle velocity measurements through application of an operator, e.g., L.
[0077] Due to differing signal-to-noise ratios of hydrophone and particle velocity data, it may be necessary to define the inversion so as to satisfy the
hydrophone data for a broader bandwidth than the particle velocity data. This may be implemented by including a frequency-dependent scaling term into the algorithm or bandpass filtering the model and data for different conjugate gradient passes either by multiplication in the frequency domain or convolution by a bandpass filter in the time domain. For example, application of L may include a bandpass filter so that, when applied, the bandwidth of particle velocity components is 25Hz to 250Hz, whereas the bandpass filter for hydrophone data is 2Hz to 250Hz.
[0078] It can also be possible to process hydrophone and particle motion data independently. The separate results may be combined afterward, for example, by selecting different temporal frequency ranges based on signal-to-noise ratio data. At low frequencies, particle velocity data may be too noisy to be of value. Therefore, it may be necessary to use a pressure-only solution for low frequencies, and make use of a combined hydrophone-particle velocity solution for higher frequencies.
[0079] A generalized weighting scheme can be implemented to vary weighting between any component (hydrophone or particle motion) depending on the receiver, time and frequency, and/or wavenumber. This weighting refers to how well the model represents the data. Typically, accelerometer data is integrated to calculate particle velocity measurements. Instead of integrating accelerometer data before wave-field separation, it is also possible to build a differentiation operator into the inversion scheme. As an alternative to the described modeling approach, other
forms of wave-field extrapolation may be used, which may include Kirchhoff, beam, wave-equation (one-way or two-way, RTM, etc.).
[0080] The above-discussed procedures and methods may be implemented in a computing device as illustrated in Figure 1 1 . Hardware, firmware, software or a combination thereof may be used to perform the various steps and operations described herein. Computing device 1 100 of Figure 1 1 is an exemplary computing structure that may be used in connection with such a system.
[0081] Exemplary computing device 1 100 suitable for performing the activities described in the exemplary embodiments may include a server 1 101 . Such a server 1 101 may include a central processor (CPU) 1 102 coupled to a random access memory (RAM) 1 104 and to a read-only memory (ROM) 1 106. ROM 1 106 may also be other types of storage media to store programs, such as programmable ROM (PROM), erasable PROM (EPROM), etc. Processor 1 102 may communicate with other internal and external components through input/output (I/O) circuitry 1 108 and bussing 1 1 10 to provide control signals and the like. Processor 1 102 carries out a variety of functions as are known in the art, as dictated by software and/or firmware instructions.
[0082] Server 1 101 may also include one or more data storage devices, including hard drives 1 1 12, CD-ROM drives 1 1 14 and other hardware capable of reading and/or storing information, such as DVD, etc. In one embodiment, software for carrying out the above-discussed steps may be stored and distributed on a CD-
ROM or DVD 1 1 16, a USB storage device 1 1 18 or other form of media capable of portably storing information. These storage media may be inserted into, and read by, devices such as CD-ROM drive 1 1 14, disk drive 1 1 12, etc. Server 1 101 may be coupled to a display 1 120, which may be any type of known display or presentation screen, such as LCD, plasma display, cathode ray tube (CRT), etc. A user input interface 1 122 is provided, including one or more user interface mechanisms such as a mouse, keyboard, microphone, touchpad, touch screen, voice-recognition system, etc.
[0083] Server 1 101 may be coupled to other devices, such as sources, detectors, etc. The server may be part of a larger network configuration as in a global area network (GAN) such as the Internet 1 128, which allows ultimate connection to various landline and/or mobile computing devices.
[0084] The disclosed exemplary embodiments provide a computing device, software instructions and a method for seismic data processing. It should be understood that this description is not intended to limit the invention. On the contrary, the exemplary embodiments are intended to cover alternatives,
modifications and equivalents, which are included in the spirit and scope of the invention as defined by the appended claims. Further, in the detailed description of the exemplary embodiments, numerous specific details are set forth in order to provide a comprehensive understanding of the claimed invention. However, one
skilled in the art would understand that various embodiments may be practiced without such specific details.
[0085] Although the features and elements of the present exemplary embodiments are described in the embodiments in particular combinations, each feature or element can be used alone without the other features and elements of the embodiments or in various combinations with or without other features and elements disclosed herein.
[0086] This written description uses examples of the subject matter disclosed to enable any person skilled in the art to practice the same, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the subject matter is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims.
Claims
1 . A method for identifying seismic traces prone to cycle-skipping in a full waveform inversion method, the method comprising:
receiving (600) recorded seismic data recorded with seismic sensors over a subsurface of interest;
selecting (602) a model that describes the subsurface;
calculating (604), based on the model and the recorded seismic data, estimated seismic data; and
choosing (606) a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data,
wherein the probabilistic measure includes at least one statistical function.
2. The method of Claim 1 , wherein the statistical function is a probability distribution function (PDF).
3. The method of Claim 2, wherein the PDF is characterized by a mean of a delay time At between traces of the recorded seismic data and traces of the estimated seismic data and its standard deviation a t.
4. The method of Claim 3, further comprising:
determining a mean frequency (/) and its associated uncertainty af by considering an amplitude spectrum of the measured seismic data as the PDF.
5. The method of Claim 2, further comprising:
calculating a cumulative distribution function (CDF), or a probability of cycle skipping Q = 1 - CDF, based on the PDF, mean delay time and its uncertainty, and mean frequency and its uncertainty.
6. The method of Claim 5, wherein a trace is considered cycle-skipped given a predetermined threshold of the CDF or Q value.
7. The method of Claim 1 , further comprising:
calculating a maximum frequency fmax based on the probabilistic measure, wherein frequencies below the maximum frequency are unlikely to have cycle- skipping.
8. The method of Claim 1 , wherein a probabilistic measure Q is derived from a PDF and is calculated for a measured time delay between the recorded and estimated seismic data.
9. The method of Claim 8, wherein a maximum frequency is given by an integral of amplitude spectrum between a minimum frequency and the maximum frequency being smaller than an inverse of twice the time delay.
10. The method of Claim 1 , further comprising:
calculating the probabilistic measure Q for a time delay At, between the recorded and estimated seismic data; and
applying weights w to a cost function C, wherein the weights are related to the probabilistic measure Q,
wherein the cost function C includes the recorded and estimated seismic data.
1 1 . A method for determining a weight based on a data selection analysis, the method comprising:
receiving (900) recorded seismic data drec recorded with seismic sensors over a subsurface of interest;
estimating (902) seismic data dest based on a starting model of the surveyed subsurface;
calculating (904) a travel-time difference (At) between the recorded data drec and the estimated data dest;
converting (906) the travel-time difference (At) to a phase difference;
selecting (914) seismic data with a phase difference between the recorded and estimated data of less than a half cycle; and
feeding (916) the selected seismic data to a full-wave inversion (FWI) model for calculating physical properties of the subsurface of interest.
12. The method of Claim 1 1 , wherein the starting model is a velocity model.
13. The method of Claim 1 1 , wherein the step of selecting is performed for each FWI iteration.
14. The method of Claim 1 1 , wherein the travel-time difference (At) is estimated using a windowed cross-correlation function.
15. The method of Claim 1 1 , further comprising:
calculating the weight based on the cross-correlation function, and smoothing the guide of selection.
16. The method of Claim 15, wherein a corresponding weight is calculated for every time sample in a seismic trace.
17. A computing device (1200) for identifying seismic traces prone to cycle- skipping in a full waveform inversion method, the computing device comprising: an interface (1208) configured to receive (600) recorded seismic data recorded with seismic sensors over a subsurface of interest; and
a processor (1202) connected to the interface (1208) and configured to, select (602) a model that describes the subsurface,
calculate (604), based on the model and the recorded seismic data, estimated seismic data, and
choose (606) a probabilistic measure that characterizes a relationship between the recorded seismic data and the estimated seismic data,
wherein the probabilistic measure includes at least one statistical function.
18. The computing device of Claim 17, wherein the statistical function is a probability distribution function (PDF).
19. A computing device (1200) for determining a weight based on a data selection analysis, the computing device (1200) comprising:
an interface (1208) configured to receive (900) recorded seismic data drec recorded with seismic sensors over a subsurface of interest; and
a processor (1202) connected to the interface (1208) and configured to,
estimate (902) seismic data dest based on a starting model of the surveyed subsurface,
calculate (904) a travel-time difference (At) between the recorded data drec and the estimated data dest,
convert (906) the travel-time difference (At) to a phase difference,
select (914) seismic data with a phase difference between the recorded and estimated data of less than a half cycle, and
feed (916) the selected seismic data to a full-wave inversion (FWI) model for calculating physical properties of the subsurface of interest.
20. The computing device of Claim 19, wherein the starting model is a velocity model.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP15735418.4A EP3092511A4 (en) | 2014-01-10 | 2015-01-09 | Device and method for mitigating cycle-skipping in full waveform inversion |
US15/107,614 US11175421B2 (en) | 2014-01-10 | 2015-01-09 | Device and method for mitigating cycle-skipping in full waveform inversion |
Applications Claiming Priority (8)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US201461925709P | 2014-01-10 | 2014-01-10 | |
US61/925,709 | 2014-01-10 | ||
US201461927194P | 2014-01-14 | 2014-01-14 | |
US61/927,194 | 2014-01-14 | ||
US201461945843P | 2014-02-28 | 2014-02-28 | |
US61/945,843 | 2014-02-28 | ||
US201462021027P | 2014-07-04 | 2014-07-04 | |
US62/021,027 | 2014-07-04 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2015106065A1 true WO2015106065A1 (en) | 2015-07-16 |
Family
ID=53524354
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US2015/010768 WO2015106065A1 (en) | 2014-01-10 | 2015-01-09 | Device and method for mitigating cycle-skipping in full waveform inversion |
Country Status (3)
Country | Link |
---|---|
US (1) | US11175421B2 (en) |
EP (1) | EP3092511A4 (en) |
WO (1) | WO2015106065A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017167191A1 (en) * | 2016-04-01 | 2017-10-05 | 中国石油天然气股份有限公司 | Method and device for processing seismic data |
CN110888159A (en) * | 2019-11-15 | 2020-03-17 | 西安理工大学 | Elastic wave full waveform inversion method based on angle decomposition and wave field separation |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10061044B2 (en) * | 2015-02-10 | 2018-08-28 | Transcend Engineering and Technology, LLC | Systems, methods, and software for detecting the presence of subterranean tunnels and tunneling activity |
EP3387465B1 (en) * | 2015-12-11 | 2021-08-04 | ION Geophysical Corporation | System and method for reconstructed wavefield inversion |
WO2017123240A1 (en) * | 2016-01-15 | 2017-07-20 | Landmark Graphics Corporation | Semblance-based anisotropy parameter estimation using isotropic depth-migrated common image gathers |
EP3559707B1 (en) | 2016-12-20 | 2021-02-17 | ION Geophysical Corporation | System and method for reconstructed wavefield imaging |
US11487036B2 (en) * | 2017-01-12 | 2022-11-01 | Cgg Services Sas | Reflection full waveform inversion methods with density and velocity models updated separately |
US11048001B2 (en) * | 2018-03-30 | 2021-06-29 | Cgg Services Sas | Methods using travel-time full waveform inversion for imaging subsurface formations with salt bodies |
DE102019117587A1 (en) * | 2019-06-28 | 2020-12-31 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung eingetragener Verein | Method, device and computer program for the detection of one or more objects in the sea floor |
CN115335729A (en) * | 2020-02-27 | 2022-11-11 | 地质探索系统公司 | Template matching full waveform inversion |
US11867856B2 (en) * | 2021-06-03 | 2024-01-09 | Saudi Arabian Oil Company | Method and system for reflection-based travel time inversion using segment dynamic image warping |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4992992A (en) | 1988-10-21 | 1991-02-12 | Western Atlas International, Inc. | Processing for seismic data from slanted cable |
US5481501A (en) * | 1993-02-19 | 1996-01-02 | Exxon Production Research Company | Method for simulating crosswell seismic data |
US5659520A (en) * | 1995-04-24 | 1997-08-19 | Sonatech, Inc. | Super short baseline navigation using phase-delay processing of spread-spectrum-coded reply signals |
US5923383A (en) * | 1996-06-27 | 1999-07-13 | Samsung Electronics Co., Ltd. | Image enhancement method using histogram equalization |
US20050028579A1 (en) * | 2002-11-18 | 2005-02-10 | Southwest Research Institute | Processing data, for improved accuracy, from device for measuring speed of sound in a gas |
US20080077337A1 (en) * | 2004-08-20 | 2008-03-27 | Takashi Yoshinaga | Electromagnetic Interference Reduction Calculation Method, Device, and Program, and Its Electronic Circuit |
US20080228402A1 (en) * | 2007-03-16 | 2008-09-18 | Ralf Ferber | Processing of Seismic Data Acquired Using Twin Over/Under Streamers |
US20090119018A1 (en) * | 2007-11-07 | 2009-05-07 | Schlumberger Technology Corporation | Subterranean formation properties prediction |
US20120176861A1 (en) * | 2011-01-12 | 2012-07-12 | Bp Corporation North America Inc. | Shot scheduling limits for seismic acquisition with simultaneous source shooting |
US20120289852A1 (en) * | 2009-12-03 | 2012-11-15 | Koninklijke Philips Electronics N.V. | Method and apparatus for estimating respiratory impedance |
US20120316791A1 (en) * | 2011-06-08 | 2012-12-13 | Chevron U.S.A. Inc. | System and method for seismic data inversion by non-linear model update |
WO2013028237A1 (en) * | 2011-08-23 | 2013-02-28 | Exxonmobil Upstream Research Company | Estimating fracture dimensions from microseismic data |
WO2013093467A1 (en) * | 2011-12-20 | 2013-06-27 | Shah Nikhil Koolesh | Method of, and apparatus for, full waveform inversion |
WO2013093468A2 (en) * | 2011-12-20 | 2013-06-27 | Shah Nikhil Koolesh | Full waveform inversion quality control method |
US8593904B2 (en) | 2010-10-14 | 2013-11-26 | Cggveritas Services Sa | Method and device to acquire seismic data |
Family Cites Families (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3714653A (en) * | 1967-12-20 | 1973-01-30 | Gen Electric | Radar system |
US3594795A (en) * | 1967-12-20 | 1971-07-20 | Gen Electric | Radar system |
IT986863B (en) * | 1972-10-31 | 1975-01-30 | Sits Soc It Telecom Siemens | RECEIVER FOR GROUP DELAY MEASUREMENT |
US4985683A (en) * | 1988-03-24 | 1991-01-15 | The Perkin-Elmer Corporation | Circuitry for measuring phase difference between two oscillatory signals |
US5287328A (en) * | 1993-01-28 | 1994-02-15 | Mobil Oil Corporation | Method for removing random source and receiver noise from seismic traces |
US5764516A (en) * | 1995-12-29 | 1998-06-09 | Atlantic Richfield Company | Method and system for surface-consistent phase and time lag correction of seismic data |
JP3617036B2 (en) * | 2001-05-22 | 2005-02-02 | ビイック株式会社 | Ground exploration device and analysis program used therefor |
US7155244B2 (en) * | 2003-08-14 | 2006-12-26 | Siemens Communications, Inc. | Precise common timing in a wireless network |
US7158899B2 (en) * | 2003-09-25 | 2007-01-02 | Logicvision, Inc. | Circuit and method for measuring jitter of high speed signals |
WO2011034870A1 (en) * | 2009-09-17 | 2011-03-24 | Board Of Regents, The University Of Texas System | Time-lapse seismic comparisons using pre-stack imaging and complex wave field comparisons to improve accuracy and detail |
US8223587B2 (en) * | 2010-03-29 | 2012-07-17 | Exxonmobil Upstream Research Company | Full wavefield inversion using time varying filters |
GB2481270B (en) * | 2011-01-26 | 2013-06-26 | Nikhil Shah | Method of, and Apparatus for, Full Waveform Inversion modelling |
US9075159B2 (en) | 2011-06-08 | 2015-07-07 | Chevron U.S.A., Inc. | System and method for seismic data inversion |
US9291735B2 (en) * | 2011-06-10 | 2016-03-22 | Globalfoundries Inc. | Probablistic subsurface modeling for improved drill control and real-time correction |
US20120320712A1 (en) * | 2011-06-20 | 2012-12-20 | Ahmed Adnan Aqrawi | Dip seismic attribute |
US9945970B1 (en) * | 2011-08-29 | 2018-04-17 | Seismic Innovations | Method and apparatus for modeling microseismic event location estimate accuracy |
GB2503640A (en) * | 2011-12-20 | 2014-01-08 | Nikhil Koolesh Shah | Quality Assurance in a Full Waveform Inversion Process |
CA2886798C (en) * | 2013-03-15 | 2022-08-30 | Chevron U.S.A. Inc. | Beam inversion by monte carlo back projection |
US20160070023A1 (en) * | 2013-04-24 | 2016-03-10 | The University Of British Columbia | A penalty method for pde-constrained optimization |
US20150088424A1 (en) * | 2013-09-20 | 2015-03-26 | Schlumberger Technology Corporation | Identifying geological formation depth structure using well log data |
GB2509223B (en) * | 2013-10-29 | 2015-03-18 | Imp Innovations Ltd | Method of, and apparatus for, full waveform inversion |
-
2015
- 2015-01-09 US US15/107,614 patent/US11175421B2/en active Active
- 2015-01-09 WO PCT/US2015/010768 patent/WO2015106065A1/en active Application Filing
- 2015-01-09 EP EP15735418.4A patent/EP3092511A4/en not_active Withdrawn
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4992992A (en) | 1988-10-21 | 1991-02-12 | Western Atlas International, Inc. | Processing for seismic data from slanted cable |
US5481501A (en) * | 1993-02-19 | 1996-01-02 | Exxon Production Research Company | Method for simulating crosswell seismic data |
US5659520A (en) * | 1995-04-24 | 1997-08-19 | Sonatech, Inc. | Super short baseline navigation using phase-delay processing of spread-spectrum-coded reply signals |
US5923383A (en) * | 1996-06-27 | 1999-07-13 | Samsung Electronics Co., Ltd. | Image enhancement method using histogram equalization |
US20050028579A1 (en) * | 2002-11-18 | 2005-02-10 | Southwest Research Institute | Processing data, for improved accuracy, from device for measuring speed of sound in a gas |
US20080077337A1 (en) * | 2004-08-20 | 2008-03-27 | Takashi Yoshinaga | Electromagnetic Interference Reduction Calculation Method, Device, and Program, and Its Electronic Circuit |
US20080228402A1 (en) * | 2007-03-16 | 2008-09-18 | Ralf Ferber | Processing of Seismic Data Acquired Using Twin Over/Under Streamers |
US20090119018A1 (en) * | 2007-11-07 | 2009-05-07 | Schlumberger Technology Corporation | Subterranean formation properties prediction |
US20120289852A1 (en) * | 2009-12-03 | 2012-11-15 | Koninklijke Philips Electronics N.V. | Method and apparatus for estimating respiratory impedance |
US8593904B2 (en) | 2010-10-14 | 2013-11-26 | Cggveritas Services Sa | Method and device to acquire seismic data |
US20120176861A1 (en) * | 2011-01-12 | 2012-07-12 | Bp Corporation North America Inc. | Shot scheduling limits for seismic acquisition with simultaneous source shooting |
US20120316791A1 (en) * | 2011-06-08 | 2012-12-13 | Chevron U.S.A. Inc. | System and method for seismic data inversion by non-linear model update |
WO2013028237A1 (en) * | 2011-08-23 | 2013-02-28 | Exxonmobil Upstream Research Company | Estimating fracture dimensions from microseismic data |
WO2013093467A1 (en) * | 2011-12-20 | 2013-06-27 | Shah Nikhil Koolesh | Method of, and apparatus for, full waveform inversion |
WO2013093468A2 (en) * | 2011-12-20 | 2013-06-27 | Shah Nikhil Koolesh | Full waveform inversion quality control method |
Non-Patent Citations (1)
Title |
---|
See also references of EP3092511A4 |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017167191A1 (en) * | 2016-04-01 | 2017-10-05 | 中国石油天然气股份有限公司 | Method and device for processing seismic data |
RU2694621C1 (en) * | 2016-04-01 | 2019-07-16 | Петрочайна Компани Лимитед | Method and apparatus for processing seismic data |
US10705238B2 (en) | 2016-04-01 | 2020-07-07 | Petrochina Company Limited | Method and apparatus for processing seismic data |
CN110888159A (en) * | 2019-11-15 | 2020-03-17 | 西安理工大学 | Elastic wave full waveform inversion method based on angle decomposition and wave field separation |
Also Published As
Publication number | Publication date |
---|---|
EP3092511A4 (en) | 2017-11-29 |
US11175421B2 (en) | 2021-11-16 |
EP3092511A1 (en) | 2016-11-16 |
US20160320505A1 (en) | 2016-11-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11175421B2 (en) | Device and method for mitigating cycle-skipping in full waveform inversion | |
US10324205B2 (en) | Device and method for full waveform inversion | |
US10338256B2 (en) | Demultiple using up/down separation of towed variable-depth streamer data | |
EP2350693B1 (en) | Continuous adaptive surface wave analysis for three-dimensional seismic data | |
US20130289879A1 (en) | Process for characterising the evolution of a reservoir | |
US10534101B2 (en) | Seismic adaptive focusing | |
US20170031045A1 (en) | Method and apparatus for modeling and separation of primaries and multiples using multi-order green's function | |
EP2270546B1 (en) | Method for dynamic aperture determination for three-dimensional surface-related multiple elimination | |
GB2391665A (en) | Producing noise-free model of distribution of heterogeneous medium in a zone | |
US9442208B2 (en) | Device and method for deghosting variable depth streamer data including particle motion data | |
US11442186B2 (en) | Source-receiver position estimation using direct arrival modeling and inversion | |
WO2013093467A1 (en) | Method of, and apparatus for, full waveform inversion | |
US9733374B2 (en) | Repeatability indicator based on shot illumination for seismic acquisition | |
EP2299296B1 (en) | Method for Combining Signals of Pressure and Particle Motion Sensors in Marine Seismic Streamers | |
US11635540B2 (en) | Methods and devices performing adaptive quadratic Wasserstein full-waveform inversion | |
US20140249757A1 (en) | Apparatus and method for determination of far-field signature from variable-depth seismic data | |
EP4080250A1 (en) | Multiple attenuation and imaging processes for recorded seismic data | |
US20170371052A1 (en) | Method for updating velocity model used for migrating data in 4d seismic data processing | |
AU2018325477B2 (en) | Source-receiver position estimation using direct arrival modeling and inversion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15735418 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15107614 Country of ref document: US |
|
REEP | Request for entry into the european phase |
Ref document number: 2015735418 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2015735418 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |