Papers by Patrick D Bangert
Day 3 Wed, February 23, 2022, Feb 21, 2022
The choice of piping classes for industrial use has important ramification on engineering project... more The choice of piping classes for industrial use has important ramification on engineering projects ranging from safety considerations to cost-material effiency and environmental footprint. Usually, suitable piping classes are found by consulting fixed tables, such as the ASME code for piping classes, together with other values like required pressure resistance and pipe temperatures, the later often being a decisive factor for choosing one piping class over another. For the pipe and flange temperatures today mostly the fluid temperature is used, while sometimes rough approximations based on fluid temperature percentages are used, or more rarely, slow, and expensive finite element calculations are performed for individual cases. In this paper we demonstrate the significant benefits of choosing correct piping classes using precisely calculated pipe temperatures and propose an AI-based approach for fast and precise calculation of these pipe temperatures based on a training set derived from a limited number of finite element calculations performed by the authors. We also analyze in detail the benefits of this approach and provide concrete examples.
CRC Press eBooks, Nov 22, 2022
Journal of Physics A: Mathematical and General, 2001
Proceedings of the ... AAAI Conference on Artificial Intelligence, Jun 26, 2023
3D deep learning is a growing field of interest due to the vast amount of information stored in 3... more 3D deep learning is a growing field of interest due to the vast amount of information stored in 3D formats. Triangular meshes are an efficient representation for irregular, nonuniform 3D objects. However, meshes are often challenging to annotate due to their high computational complexity. Therefore, it is desirable to train segmentation networks with limitedlabeled data. Self-supervised learning (SSL), a form of unsupervised representation learning, is a growing alternative to fully-supervised learning which can decrease the burden of supervision for training. Specifically, contrastive learning (CL), a form of SSL, has recently been explored to solve limited-labeled data tasks. We propose SSL-MeshCNN, a CL method for pre-training CNNs for mesh segmentation. We take inspiration from prior CL frameworks to design a novel CL algorithm specialized for meshes. Our preliminary experiments show promising results in reducing the heavy labeled data requirement needed for mesh segmentation by at least 33%.
arXiv (Cornell University), Aug 8, 2022
3D deep learning is a growing field of interest due to the vast amount of information stored in 3... more 3D deep learning is a growing field of interest due to the vast amount of information stored in 3D formats. Triangular meshes are an efficient representation for irregular, nonuniform 3D objects. However, meshes are often challenging to annotate due to their high geometrical complexity. Specifically, creating segmentation masks for meshes is tedious and timeconsuming. Therefore, it is desirable to train segmentation networks with limited-labeled data. Self-supervised learning (SSL), a form of unsupervised representation learning, is a growing alternative to fully-supervised learning which can decrease the burden of supervision for training. We propose SSL-MeshCNN, a self-supervised contrastive learning method for pre-training CNNs for mesh segmentation. We take inspiration from traditional contrastive learning frameworks to design a novel contrastive learning algorithm specifically for meshes. Our preliminary experiments show promising results in reducing the heavy labeled data requirement needed for mesh segmentation by at least 33%. * This work was conducted during an internship with Samsung SDS Research America. A. Haque is currently affiliated with UC Berkeley.
Nuclear Instruments and Methods in Physics Research, Jun 1, 1998
In the production of ultracold neutrons (UCN) by scattering cold neutrons in superfluid 4 He or c... more In the production of ultracold neutrons (UCN) by scattering cold neutrons in superfluid 4 He or cold solid D 2 ("superthermal" processes), the UCN production rate is proportional to the spatial density of cold neutrons in the material. For a superthermal source fed from a cold neutron guide, we show that the rate of UCN production can be increased by up to a factor of 2.5 for realistic substances by enclosing the source in a cavity made of a material with a high cold neutron albedo.
Biomedical Engineering
Multiomic analysis comprises genomics, proteomics, and metabolomics leads to meaningful insights ... more Multiomic analysis comprises genomics, proteomics, and metabolomics leads to meaningful insights but necessitates sifting through voluminous amounts of complex data. Proteomics in particular focuses on the end product of gene expression – i.e., proteins. The mass spectrometric approach has proven to be a workhorse for the qualitative and quantitative study of protein interactions as well as post-translational modifications (PTMs). A key component of mass spectrometry (MS) is spectral data analysis, which is complex and has many challenges as it involves identifying patterns across a multitude of spectra in combination with the meta-data related to the origin of the spectrum. Artificial Intelligence (AI) along with Machine Learning (ML), and Deep Learning (DL) algorithms have gained more attention lately for analyzing the complex spectral data to identify patterns and to create networks of value for biomarker discovery. In this chapter, we discuss the nature of MS proteomic data, the...
The invention relates to a grinder robot for a vessel hull comprising a reactor equipped with a m... more The invention relates to a grinder robot for a vessel hull comprising a reactor equipped with a movable abrasive grinding head, which is attached to an actuator, connecting the grinding head with a frame and can move in at least two perpendicular directions relative to the frame within a machining space. According to the invention an electronic control unit is provided which is signally connected to the actuator for controlling the movement of the grinding head along the at least two directions, comprising the control unit comprises: a filtering module, which is designed for the removal of scan data that are outside of the ship hull.
Machine Learning and Data Science in the Oil and Gas Industry, 2021
Abstract This chapter will attempt to provide an overview over some of the practical applications... more Abstract This chapter will attempt to provide an overview over some of the practical applications that machine learning has found in oil and gas. The aim of the chapter is twofold: First, it is to show that there are many applications that are realistic and have been carried out on real-world assets, that is, machine learning is not a dream. Second, the status of machine learning in oil and gas is in its early days as the applications are specialized and localized. It must be stated clearly that most of the studies done, have been done at universities and that the applications fully deployed in commercial companies are the exception. This chapter makes no attempt at being complete or even representative of the work done. It just provides many starting points for research on use cases and presents an overview. There are some use cases that attract a vast number of papers and this chapter will present such use cases with just one or a few exemplary papers chosen at random.
A "black box" simulation of several shallow-water oil wells producing to a single platform reveal... more A "black box" simulation of several shallow-water oil wells producing to a single platform revealed that the field's maximum production rate could be increased by 5%.
Machine Learning and Data Science in the Oil and Gas Industry, 2021
Abstract Approximately 20% of all oil wells in the world use a beam pump to raise crude oil to th... more Abstract Approximately 20% of all oil wells in the world use a beam pump to raise crude oil to the surface. The proper maintenance of these pumps is thus an important issue in oilfield operations. We wish to know, preferably before the failure, what is wrong with the pump. Maintenance issues on the downhole part of a beam pump can be reliably diagnosed from a plot of the displacement and load on the traveling valve; a diagram known as a dynamometer card. This chapter shows that this analysis can be fully automated using machine learning techniques that teach themselves to recognize various classes of damage in advance of the failure. We use a dataset of 35292 sample cards drawn from 299 beam pumps in the Bahrain oilfield. We can detect 11 different damage classes from each other and from the normal class with an accuracy of 99.9%. This high accuracy makes it possible to automatically diagnose beam pumps in real-time and for the maintenance crew to focus on fixing pumps instead of monitoring them, which increases overall oil yield and decreases environmental impact.
Most of the effort in a data science project lies in getting a clean, representative, informative... more Most of the effort in a data science project lies in getting a clean, representative, informative dataset. This chapter discusses all the relevant steps in getting to this point. A brief discussion of measuring and storing data in control systems and historians starts us off and we observe along the way that all data is uncertain to some degree. Time-series have an inherent time scale and may be correlated with each other. These relationships can be encoded in a mathematical representation of the process that made the data, i.e., the model. The twin concepts of representation and significance of the data are discussed so that the data suits the task at hand. Outliers are removed next and we discuss how to judge the goodness of a model against the data. Next, we discuss feature engineering to add domain knowledge and dimensionality reduction to keep the size of the dataset low while maintaining its information content. The chapter concludes with a practical checklist to see if the da...
It is often difficult or expensive to measure a physical quantity with a regular sensor. Many sen... more It is often difficult or expensive to measure a physical quantity with a regular sensor. Many sensors are fragile, especially in the harsh environments common in the oil and gas industry. The cost, both financial and in effort, to install and maintain a sensor over the long term can be entirely saved if it is possible to calculate the desired value, instead of measuring it. This calculation is called a soft sensor and can be based on known formulas or a machine learned model. In this chapter, we investigate the creation of a soft sensor for NOx and SOx emissions in a combined heat and power plant in an oilfield. We find that the calculation is as accurate as a physical measurement, available always, and deployable in multiple locations without additional costs.
Wind power plants are uniquely exposed to various damage mechanisms from exposure to weather. The... more Wind power plants are uniquely exposed to various damage mechanisms from exposure to weather. These have serious impact upon the lifetime cost of the plants and the power production that can be expected from them. Nonetheless, the damages are visible in the vibration spectra well before they lead to failures. By making use of the information, predictive maintenance can be enabled. This chapter analyzes the ways in which this is done and illustrates that failures can be successfully forecasted on turbine blades, rotors, and generators.
CHAPTER 6. COMBINATORICS-II Case 2: |S| ≥ 3. If any pair in S are strangers then those two along ... more CHAPTER 6. COMBINATORICS-II Case 2: |S| ≥ 3. If any pair in S are strangers then those two along with a are three mutual strangers. Else S becomes a set of mutual friends of size at least 3. 4. Let {x 1 ,. .. , x 9 } ⊆ N with 9 i=1 x i = 30. Then, prove that there exist i, j, k ∈ {1, 2,. .. , 9} with x i + x j + x k ≥ 12. Ans: Note that 9 i=1 x i 9 = 30 9 = 3 + 3 9. Now use PHP to conclude that there are at least 3 x i 's that are ≥ 4. Hence, the required result follows. 5. Each point of the plane is colored red or blue, then prove that there exist two points of the same color which are at a distance of 1 unit. Ans: Take a point, say P. Draw a unit circle with P as the center. If all the points on the circumference have the same color then we are done. Else, the circumference contains a point which has the same color as that of P. 6. If 7 points are chosen inside or on the unit circle, then there is a pair of points which are at a distance at most 1. Ans: Divide the circle into 6 equal sectors by drawing radii so that angle between two consecutive radii is π/3. By PHP there is a sector containing at least two points. The distance between these two points is at most 1. 7. If n + 1 integers are selected from {1, 2,. .. , 2n}, then there are two, where one of them divides the other. Ans: Each number has the form 2 k s, where s = 2m + 1 is an odd number. There are n odd numbers. If we select n + 1 numbers from S, by PHP some two of them (say, x, y) have the same odd part, that is, x = 2 i s and y = 2 j s. If i ≤ j, then x|y, otherwise y|x. 8. Given any n integers, n ≥ 1012 integers, prove that there is a pair that either differ by, or sum to, a multiple of 2021. Is this true if we replace 1012 by 1011? Ans: Consider some 1012 integers out of the given ones, say, n 1 , n 2 ,. .. , n 1012. Write S = {n 1 − n k , n 1 + n k : k = 2,. .. , 1012}. Then, |S| = 2022 and hence, at least two of them will have the same remainder when divided by 2021. Then, consider their difference. The question in the second part has negative answer. For, consider {0, 1, 2,. .. , 1010}. 9. Prove that there exist two powers of 3 whose difference is divisible by 2021.
Uploads
Papers by Patrick D Bangert