Search

Seismic data contain information about elastic properties and the structural patterns of the subsurface. However, before we get an image of the subsurface, seismic data goes through many processing steps that consist of applying theory of wave propagation and mathematical and statistical signal processing. In this thesis, we have investigated breakthroughs within artificial intelligence for seismic wavefield reconstruction. In the first paper, we developed a method that is suited for marine collection where one can utilize similarities between acquisition domains to train the machine learning model. In the second paper, we utilize methodology from image inpainting combined with inverse theory to fit the reconstruction problem in the offset domain, within an unsupervised deep learning framework. In the final paper, we illustrate how one can demigrate field data to generate suitable training data for interpolation and regularization of offset classes. This synthesized training data can then be used to create the prediction models. In all three papers, we have documented an improvement compared to existing "state-of-the-art" methods from the industry.

BACKGROUND: Discrimination between recurrent glioblastoma and therapy-related changes with conventional magnetic resonance imaging (MRI) is challenging. There is increasing evidence for the clinical value of positron emission tomography (PET) with O-(2-18F-fluoroethyl)-L-tyrosine (FET) in addition to MRI in cases of ambiguous findings in the setting of posttreatment care. The diagnostic accuracy, however, varies among the published studies and is influenced by the composition of the population with different glioma subtypes/grades, lack of histological confirmation, and differences in data processing. Herein, we evaluated the diagnostic accuracy of FET PET scans in glioblastoma patients. MATERIAL AND METHODS: One hundred thirty-seven consecutive patients who had undergone 150 FET PET were reviewed retrospectively. Inclusion criteria were 1) histologically-proven glioblastoma; 2) previous surgery followed by oncological treatment consisting of standard radiochemotherapy or - if at second recurrence - chemotherapy; 3) unexplained constant or increasing contrast-enhancing (CE) lesions on T1, or non-CE lesions on T2 FLAIR MRI later than 6 months after radiotherapy, where the differentiation between disease recurrence or therapy-related changes was uncertain; 4) FET PET for supplementary evaluation; 5) a histological evaluation following surgery less than 3 months after FET PET, or MRI follow-up. FET PET scans were performed as 20-minute static PET/CT acquisitions and evaluated co-registered to T1 post-contrast MRI with measurement of maximum and mean tumor-to-brain ratios (TBR(max,) TBR(mean)). Receiver operating characteristics (ROC) analysis was used to determine the optimal threshold of FET parameters. The prognostic influence of FET parameters on overall survival (OS) was investigated with the Cox proportional hazards model. RESULTS: The median time interval from radiotherapy until the radiological progression was 13 months. One hundred twenty-six PET scans demonstrated FET uptake of varying intensity. Surgical interventions were performed following 88 PET scans, while 62 PET scans were evaluated by clinical or MRI follow-up, resulting in 131 glioblastoma recurrences and 19 therapy-related changes. ROC analysis yielded the thresholds of 2.0 for TBR(max) and 1.8 for TBR(mean) for differentiation between recurrent disease and therapy-related changes with the best performance of TBR(max) with a sensitivity of 96% and a specificity of 100% (p < 0.0001). Using this threshold, 145 of 150 PET scans of glioblastoma recurrence or therapy-related changes were accurately classified. Multivariate survival analysis showed that TBR(max) > 2.0 predicted independently a shorter OS in patients, who did not undergo subsequent therapy following PET (p = 0.002). CONCLUSION: FET PET is a powerful noninvasive tool to distinguish recurrent glioblastoma from therapy-related changes.

This paper presents the results from a series of physical modeling and three-dimensional finite 9 element (FE) analyses, which examined the uniaxial vertical capacity of suction caissons for 10 offshore wind turbines. The experiments were carried out in quartz sand and involved monotonic 11 application of vertical load. It was found that the drained capacity of suction caissons is 12 dependent on embedment ratio. On the other hand, predictions from conventional semi-empirical 13 depth factors were found to somewhat underestimate when applied to rough foundations. On the 14 basis of the tests and FE analyses, new expressions for depth factor of shallow foundations have 15 been validated for embedment ratios (aspect ratios) up to unity, calibrating the fitting parameters 16 by employing data from a range of soil profiles.

The purpose of this master thesis is to develop data-driven mathematical models to “simulate” an equipment lifecycle over a period of time to reveal the associated effects like survival curve and hazard rate. The failure frequency or the hazard rate is input to a cost/benefit analysis with cost of failure and cost of maintenance to find the optimal maintenance interval. Furthermore a baseline preventive maintenance program made with Reliability Centred Maintenance methodology is compared with the PM-program optimized with use of the data-driven methodology. The model is tested and used with a sample size of 100, random sampled over a period of 10 years and uses OREDA failure mode dataset with Mean failure rates as input. Survival Analysis is developed with use of Lifelines resulting into estimated survival (reliability) function with and the Kaplan Meier estimate and estimated Hazard rate with the Nelson Aalen estimate. The study shows potentials of cost savings using data-driven modelling; however the most beneficial is that the data-driven modelling results into a decision basis for cost/benefit analysis for optimizing maintenance. Decision basis support like chance of asset survival for a given time interval, MTTF (Mean Time To Failure) and hazard rate. Last but not least recommendations for further work are discussed.

A series of comprehensive research and development projects has validated the novel concept of the bucket foundation as an alternative to piled foundations, and bucket foundations are rapidly becoming a viable option for offshore wind farms and marginal fields.Within the last decade, experimental studies have indicated the possibility of defining foundation response using plasticity theory. Results of multiple loading tests addressing the effect of embedment on the strain-hardening behavior of shallow bucket foundations under combined loading are reported. The kinematic mechanisms accompanying pre-failure are presented. It is argued that the drained capacity of offshore bucket foundations and the ratio of plastic increments are largely influenced by embedment depth and the preload ratio V/Vpeak.

A new concept with respect to offshore wind turbines, the bucket foundation, is known in the oil and gas industry, although the load conditions for wind turbines are significantly different. An extensive number of loading tests with small-scale bucket foundations subjected to combined loading were carried out in the geotechnical laboratory at Aalborg University in Aalborg, Denmark. Tests were performed on buckets of various sizes, embedment ratios, and load paths on saturated dense Aalborg University Sand No. 1. The experimental results were used to evaluate the behavior of the bucket foundations in accordance with the macromodel approach. An expression to describe the combined capacity of bucket foundations was developed by calibrating the failure criteria for bucket foundations.

The bucket foundation is a new type of foundation, which has been developed over the past 3 years and today it is used for a Vestas V90-3.0 MW offshore wind turbine as a prototype. Ibsen et al.(2003). The bucket foundation is believed to have great potential in the future offshore wind parks as the wind turbines are getting larger and more effective. The behaviour of bucket foundations located in saturated dense sand has been investigated by use of model tests. One of the methods to predict the response of a full size bucket foundation is the Finite Element Method, FEM. This method has been used to back calculate the response of the model tests in the laboratory. It has been shown that a three-dimensional Finite Element model is a useful method to predict the behaviour of bucket foundations for offshore wind turbines installed in saturated sand. The bucket foundation is a new type of foundation, which has been developed over the past 3 years and today it is used for a Vestas V90-3.0 MW offshore wind turbine as a prototype. Ibsen et al.(2003). The bucket foundation is believed to have great potential in the future offshore wind parks as the wind turbines are getting larger and more effective. The behaviour of bucket foundations located in saturated dense sand has been investigated by use of model tests. One of the methods to predict the response of a full size bucket foundation is the Finite Element Method, FEM. This method has been used to back calculate the response of the model tests in the laboratory. It has been shown that a three-dimensional Finite Element model is a useful method to predict the behaviour of bucket foundations for offshore wind turbines installed in saturated sand.