IET AI Lab is the AI Baker Hughes Research Laboratory. See also Baker Hughes AI

The AI Team's mission, which directly derives from the company's vision, is to improve safety, cleanliness and efficiency of energy with the utilization of intelligent technologies.

We want to accomplish this mission by following two parallel threads, the operational (first row in the graphic above) and the cultural (second row).

The goal of the operational thread is to demonstrate, on the basis of real projects, the AI impact on all of the company's activities in terms of productivity and revenues increase. To achieve this goal, we have defined two different sets of programs: the Lines of Action, which collect application projects with a well defined, internal customer, and the Enablers, which collect the development of tools and platforms aimed at increasing the execution speed of AI projects.

The goal of the cultural thread is to foster and accelerate a cultural change along the dimensions of agility, collaboration and competence.

More than 20 projects delivered in productions. Reinforcement Learning applied to controls, Document Understanding via NLP, Automatic Visual Inspection via Computer Vision, Data Forecasting using ML.

The goal is to automate inspection activities on a mineral lube oil skid used to lubricate gas turbines. The robot navigates autonomously around the skid and performs inspections on items such as analogic indicators, digital displays, sight glasses, level gauges, and electric motors using cameras, thermal cameras and microphones. Data is collected over routine missions to allow the development and training of additional computer vision and machine learning models.

Identify potential anomalies using semi supervised approach.

Accelerate the aerodynamic design process using AI. Aerodynamic designers use CFD to predict the performance of a new design or to find the effect of a geometry modification on an existing design. Defining new geometries, creating a mesh, launching a set of CFD runs (speedline) and postprocessing results requires significant human and computational resources. We are investigating different approaches (supervised and semi-supervised learning) to explore the design space and automatically find optimal designs, without the need to run one or more speedlines for each new design.

We use LLM prompting to extract alloy formula from simple text.

1. 50+ Publications
2. 5+ Patents
3. A team of 40+ peoples
4. 10+ PhD Graduated

• Online Sequence-Based Deep Learning Approach for Metallic Debossed and Embossed Turbomachinery Blade Text Recognition Application - A. Youssef, P. Mishra, M. Vitale, G. Schillaci, G. Veneri, A. Bettini, G. Anatriello, M. Burbui, F. Ceccherini - International Petroleum Technology Conference, Dhahran, Saudi Arabia, February 2024 link

• Anomaly Detection of Sensor Measurements During a Turbo-Machine Prototype Testing - An Integrated ML Ops, Continual Learning Architecture - Somasundaram Palaniappan; Giacomo Veneri; Valentina Gori; Tommaso Pratelli; Valeria Ballarini - International Petroleum Technology Conference, Dhahran, Saudi Arabia, February 2024 - link

• Controllable Image Synthesis of Industrial Data Using Stable Diffusion - Gabriele Valvano, Antonino Agostino, Giovanni De Magistris, Antonino Graziano, Giacomo Veneri - 2024 - Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision link

• Combining Thermodynamics-based Model of the Centrifugal Compressors and Active Machine Learning for Enhanced Industrial Design Optimization - S Ghiasi, G Pazzi, C Del Grosso, G De Magistris, G Veneri - 1st workshop on Synergy of Scientific and Machine Learning Modeling, SynS - 2023 link

• Sensor virtualization for anomaly detection of turbo-machinery sensors - An industrial application - S Shetty, V Gori, G Veneri - 9th International conference on Time Series and Forecasting - ITISE2023 2023 link

• Document Layout Analysis with Variational Autoencoders: An Industrial Application - A Youssef, G Valvano, G Veneri - ISMIS 2022. Lecture Notes in Computer Science 13515 2022 link

• Continual Learning for anomaly detection on turbomachinery prototypes - A real application - V Gori, G Veneri, V Ballarini - 2022 IEEE Congress on Evolutionary Computation (CEC), 2022, 1-7 2 2022 link

• Deep Surrogate of Modular Multi Pump using Active Learning - M Murugesan, K Goyal, L Barriere, M Pasquotti, G Veneri, G De Magistris - Adaptive Experimental Design and Active Learning in the Real World - ICML 2022 1 2022 link - arxiv

• DANNTe: a case study of a turbo-machinery sensor virtualization under domain shift - L Strazzera, V Gori, G Veneri - Distribution shifts: connecting methods and applications (DistShift) - NEURIPS 2 2021 arxiv

• Learning to Identify Drilling Defects in Turbine Blades with Single Stage Detectors - A Panizza, ST Stefanek, S Melacci, G Veneri, M Gori - Machine Learning for Engineering Modeling, Simulation, and Design Workshop 2020 arxiv

  See also papers

• cocohelper

• NLP TK : https://github.com/AILAB-bh/ailab-nlptk
• CV TK : https://github.com/AILAB-bh/ailab-cvtk
• XAI and TS TK: https://github.com/AILAB-bh/ailab-xai
• DAI TK: https://github.com/AILAB-bh/ailab-dai
• ROBOTIC TK : https://github.com/AILAB-bh/ailab-RandC
• MLOPS TK : https://github.com/AILAB-bh/ailab-library

• : BH IET copyright

• : part of ailab tk

• : made with ailab tk

• : dependencies verified

• : Clean Code

• : ALPHA VERSION

• : DEV VERSION

• : POC VERSION

• : PROD VERSION