--> An IIoT Platform for Agile Development and Deployment of Data-Driven Solutions in E&P

2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

An IIoT Platform for Agile Development and Deployment of Data-Driven Solutions in E&P

Abstract

In the recent decade, upstream companies’ investment in digital transformation and data-driven analytics increased rapidly. Machine learning is expected to bring enhanced insights into E&P workflows that are traditionally based on physics and engineering knowledge. However, the majority of data science projects did not go beyond research papers and few were put in constant operation in actual business processes. Data scientists in upstream companies spend most their time on their local computing environments, which are isolated from real-time data and real users. This gap obstructs continuous testing and improving of data science solutions in real-world scenarios.

Agile methodology is a widely-used product development workflow in Silicon Valley. It emphasizes the importance of continuous feedback from end users and fast cycles of product operationalization. We built an industrial IoT platform that aims at enabling agile development and deployment of data science solutions in E&P. The platform allows data scientists to operationalize machine learning solutions with minimal IT supports, and it accelerates the feedback-improvement iterations between end users and data scientists effectively.

Core components of this integrated platform includes:

1. Edge agent: a software installed on assets (e.g. equipements, sensors, IoT devices) that standardizes real-time data streams. It also hosts models of edge computing which are usually for smart processing of raw data.

2. Cloud-based model hub: the center of entire system that connects data, models, and applications. The hub automates enterprise-scale model management (e.g. resourcing, CI/CD, version management). Containerization technology plays a key role here.

3. Model deployment software: a tool that enables data scientists to publish models from local environment to model hub and/or to edge agent. Data scientists may operationalize a new solution such that it is ready to end users in just a few minutes.

4. End-user application: it includes flexible connector to common BI tools, a web-based dashboard builder, and an automated web-based interface renderer.

We tested the platform with our industrial partners in real-world E&P environments. We will highlight a couple of geology-related cases: one on automated well log interpretation, and the other on machine learning assisted seismic imaging.