Patent EP0731363A2 successfully refreshed from Google Patents.
EP0731363A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Neural network interpretation of aeromagnetic data
Assignee: Texaco Development Corp
•
Filing: 1992-09-17
•
Granted: —
•
Expires: 2012-09-17
Abstract
Abstract A method for determining depth to basement from aeromagnetic data utilizes neural networks (10) to automate the laborious process of profile interpretation. The neural networks provide consistency, accuracy and overall quality without bias of interpretation. A network is provided having input (12), intermediate (14) and output (16) layers. A two-dimensional profile magnetic basement model is formed, reprocessed aeromagnetic survey profiles are windowed for input to the network, the network is trained by weight adjustment using profiles from known basement structures, the trained network is tested and then applied to aeromagnetic profiles over an unknown earth formation. <IMAGE>
Inventors
- 👤 Robert Frederick Moll
- 👤 William Chelsie Pearson
- 👤 John Robert Rogers
- 👤 Jacky Muri Wiener
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-10-26 23:13
CPC Classifications
G; G01; G01V; G01V3/00; G01V3/38; G01V3/15; G01V3/165
Claims Analysis
Total Claims: 16
Claim 1
Independent Claim
58 words
A neural network for studying structural anomalies along aeromagnetic traces of earth formations, said network characterised by: an input layer (12) having a plurality of elements; at least one intermediate layer (14) having a plurality of elements; and an output layer (16) having a plurality of elements and adapted for training to distinguish between suprabasement and intrabasement structures.
Claim 2
Independent Claim
140 words
A method for studying structural anomalies along aeromagnetic traces of earth formations, characterised by the steps of: providing a neural network having at least input, intermediate and output layers; forming at least one two-dimensional profile magnetic basement structural model, the or each said model having a range of geometric dimensions approximating an area to be studied; reprocessing aeromagnetic survey profiles to remove erroneous readings and the International Geomagnetic Reference Field; windowing said profiles for input into the network; training the neural network through weight adjustment using computer generated magnetic profiles from basement structures representative of those anticipated to be found in the basin of interest; testing the trained network on aeromagnetic profiles where basement structures are already known from other means to validate the network; and applying the network to recorded aeromagnetic profiles over an unknown portion of the basin.
Claim 3
Dependent Claim
43 words
A method according to Claim 2 further characterised by the step of adjusting node weights during the learning process to minimize the error between network prediction and known models according to the Delta Rule of learning and back-propagation as applicable to neural networks.
Claim 4
Dependent Claim
32 words
A method according to Claim 2 characterised in that said windowing is accomplished by: lagging the profile response from the first step with samples spaced symmetrically to make a window of values.
Claim 5
Dependent Claim
46 words
A method according to claim 2 further characterised by the step of: creating a cumulative back-propagation network of a plurality of processing elements in the input layer, a plurality of elements in at least one hidden layer, and at least one element in an output layer.
Claim 6
Dependent Claim
27 words
A method according to Claim 4 further characterised by: affixing a non-linear transfer function to the processing elements in order to achieve optimum learning through weight adjustment.
Claim 7
Dependent Claim
17 words
A method according to Claim 6 characterised in that said non-linear transfer function is a hyperbolic tangent.
Claim 8
Dependent Claim
27 words
A method according to Claim 2 further characterised by: affixing a linear transfer function to the processing elements in order to achieve optimum learning through weight adjustment.
Claim 9
Independent Claim
39 words
A method according to any one of Claims 2 to 8 further characterised by: allowing the network to learn for a plurality of iterations, then jogging the weights and resuming learning for at least twice the number of iterations.
Claim 10
Independent Claim
25 words
A method according to any one of claims 2 to 9 further characterised by: testing the network by calculating structural values from the input data.
Claim 11
Independent Claim
54 words
A method according to any one of Claims 2 to 10 used for basement depth determination, or to locate basement structures, or to locate basement faults, or to determine vertical relief of basement structures, or to determine separation of structural and compositional anomalies on aeromagnetic profiles, or for aeromagnetic profile interpretation for basement mapping.
Claim 12
Independent Claim
91 words
A method according to any one of Claims 2 to 11 further characterised by the steps of: creating a cumulative back-propagation network of at least one processing element in the input layer, at least one element in at least one hidden layer, and at least one element in an output layer; allotting a transfer function to the processing elements in order to achieve optimum learning; and allowing the network to learn for a plurality of iterations, then jogging the weights and resuming learning for a least twice the number of iterations.
Claim 13
Dependent Claim
57 words
A method according to Claim 12 characterised in that a processing element can have many input paths coming into it and whose value is the sum of the incoming neuron values times their respective connection weights: I j = ΣW ji X i where I j is termed the Internal Activation of the j th processing element.
Claim 14
Dependent Claim
34 words
A method according to Claim 13 characterised in that the interval activation of this processing element is modified by a linear transfer function which then becomes input to the next layer of processing elements.
Claim 15
Independent Claim
170 words
A method according to any one of Claims 12 to 14 characterised in that the step by which the matrix of weights are determined or adjusted in response to an input is referred to as the learning rule and, back-propagation, the responsibility for reducing output error is shared among all of the connection weights, said error first being calculated at the output layer by the difference between the desired output and the actual output, then multiplying this difference by the derivative of the transfer function: E ( out ) k = F' ( I ( out ) k )*( D k - O k ) and using a sum of products to the previous layer, the previous proportional errors are then assigned proportional errors computed in a fashion similar to a feed forward sum of products: E ( s ) j = F' ( I ( s ) j )*Σ k ( E ( s +1) k * W ( s +1) kj )
Claim 16
Dependent Claim
52 words
A method according to Claim 15 characterised in that the recursive process back-propagates the error throughout the network and finally the weights are adjusted with the following formula: Δ W ji = lcoef * E (s ) j * X ( s -1) i where: lcoef = learning coefficient
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.