This work introduces the Cycle Orchestrator, a microservices infrastructure to structure and manage workflows related to heterogeneous data from the O&G domain. Through a knowledge-based perspective, it leverages reasoning, explainability and collaboration among stakeholders.
Domain and requirements. In the natural resources domain, particularly in the oil and gas (O&G) industry, seismic data interpretation is key in exploration processes to identify geological structures in the subsurface, allowing experts to detect patterns and correlate geological factors by exploring different data sources. Commonly, this practice involves processing massive amounts of data through diverse techniques, aiming at detecting geological structures, enhancing information, correcting potential inconsistencies in the data acquisition process, and other purposes. An increasing number of works in the literature have been proposed applying Machine Learning (ML) workflows to support aspects of such processing. To systematically model geological exploration processes that apply complex data processing pipelines, allowing other stakeholders to collaborate and consume experiments’ results, a holistic perspective is required. In this sense, we conceptualized and developed the Cycle Orchestrator, a knowledge-based workflow management system (WfMS) to support and operationalize the whole lifecycle of ML and general-purpose workflows. Including specification, setup, execution and provenance data management of such workflows. It was conceived within the O&G domain, primarily to support exploration use cases that apply cyclic ML workflows. Streams of tasks that can yield improved results through a chain of execution iterations. These workflows are associated to particular types of data sources (e.g. pre-stack and post-stack seismic data). The considered use cases comprised unsupervised ML pipelines that produce (train) new models and reuse pre-trained models and weights against new datasets, improving the quality by cyclic evolution. In this context, the orchestration involves the definition of what model and version should be applied to analyze a specific data source, as required by particular workflows applied in O&G exploration processes.
Knowledge-based Workflow Management. The Cycle Orchestrator takes advantage
of the Hyperknowledge [
Cycle Orchestrator
API
Orchestrator on Specification Controler iiitfccape lrde SpePcaifrisceartion loSw naH fr k o W
Workflow Builder
aligned through a domain ontology to segments of multimodal content. Information is
represented in the Hyperknowledge Base, a hybrid storage solution that uses a direct
hyperlinked knowledge graph to maintain all information about workflow execution
plans and provenance data stored in a knowledge base. The proposed modeling adheres
to the MLWfM ontology [
Controler Service itxceuon lre E(AxepcaucthioenAEirnflgoiwn)e lfrkooEwW adnH OauntpduLtoDgasta ProvDeantaance Hyperknowledge Base e g a iLne lrde fkow naH lr o W
Users interact with the system through a REST API and a web UI for curating and
querying information, named Knowledge Explorer System (KES) [
By integrating workflows’ lifecycles in a common representation, our approach promotes knowledge production, consumption and curation in the O&G domain. Enabling industry experts to design exploration processes holistically, connecting heterogenous data processing, ontologies and stakeholders.