ADVERTISEMENT

Semantic Framework Aligns Real-Time Drilling-Management and Control Applications

You have access to this full article to experience the outstanding content available to SPE members and JPT subscribers.

To ensure continued access to JPT's content, please Sign In, JOIN SPE, or Subscribe to JPT

Real-time signals exchanged during drilling operations are in constant evolution, provided by multiple stakeholders that have distinct perspectives on the drilling process. To achieve improved drilling management and control, the seamless exchange of real-time drilling data, without the necessity of any human intervention to configure software applications, is desirable. The complete paper presents a drilling semantic framework that allows software solutions to achieve automatic and versatile self-configuration. However, the data-exchange performances are compatible with the requirements of high-end applications involved in the management and control of drilling operations.

Motivation

Many data providers are involved in the well-construction process. Their number and role change throughout the drilling process as a function of the different activities that are executed. A typical activity lasts between 1 day and 2 weeks, and, consequently, the constellation of real-time data providers, in the most extreme scenarios, may evolve at the same pace. Furthermore, the drilling system is not static in nature. Typically, new downhole measurement tools are used for every drillstem that is run in hole.

Either because of a different setup of data provider companies or as a result of a modification of the drilling system, the number and nature of ­real-time drilling signals that are active during the well-construction process change almost daily. For any application that needs access to such real-time data feeds, a manual reconfiguration of the system inputs is both cumbersome and a potential source of error. It, therefore, is desirable that applications can automatically and seamlessly expose, discover, and choose the signals that are relevant for the management and control of the drilling process.

In the complete paper, to illustrate how automatic discovery and interpretation of the semantics of signals can provide seamless interoperability and improve the quality of service, the authors proceed step by step through an example centered on anomaly detection and characterization. The data consumer application is a data-driven system that detects abnormal hookloads and attempts to characterize the cause of the anomaly. This synopsis will discuss the theoretical underpinnings of the authors’ approach rather than outline the example detailed in the complete paper.

Class Hierarchies To Capture Specialization and Generalization

In computer science, and more specifically in object-oriented programming, entities that share the same properties and behavior (i.e., functions or procedures, also called methods) are described by a class. A class defines properties and methods that are common to a group of objects. A realization of a class is called an instance of the class and can have its own specific values for each of its properties. It is possible to define subclasses that inherit properties and behavior from the parent classes, but that, in addition, can define new or specialized versions of the properties and methods compared with the parent class. Most object-oriented languages support single inheritance (i.e., a given class can only be a subclass of a single parent class). This defines a directed hierarchical structure of classes, in which classes deeper in the tree structure are more specialized than the shallower ones, or, to put it differently, shallower classes are generalizations of the deeper ones. A class hierarchy is well-suited to describe a decomposition of a domain that needs to be adjusted rarely.

Strongly Typed Relations To Define a Semantic Network

Single inheritance can lead quickly to the necessity of duplicating properties and methods in a class hierarchy when a similar behavior shall be described for classes that do not directly inherit from the same parent class. For instance, a specific behavioral property could be the dependence of the physical quantity on pressure and temperature. Such a behavior is relevant for mass densities of materials; rheological behavior of viscous fluids; and specific heat capacity and thermal conductivity of solids, liquids, or gas. Yet each of these physical values is associated with classes that are on distinctive branches of the physical property class hierarchy. On the other hand, defining such a property at a common ancestor level in the class hierarchy, in order to benefit from inheritance, would require disabling the inheritance for any other physical values that do not depend on pressure and temperature conditions.

To solve this problem, some programming languages support multiple inheritance (i.e., a class can inherit from several classes), allowing the reuse of behavioral code across classes. Other languages solve this problem with the notion of interfaces. An interface only specifies that a set of properties and methods shall be implemented for classes that inherit from that interface. However, implementations of these interfaces are not inherited and genericity must be used if implementation is to be shared across multiple classes. Because different programming languages have various approaches to solving the generic problem of multiple inheritance, the portability across multiple programming languages of a semantic definition of drilling real-time signals could easily be restricted.

A simple way to address both the problem of multiple inheritance and dynamic semantic description is to use the concept of a semantic network. The authors define two top class hierarchies, one named Node and one named Relation. A Relation has two properties of type Node: “from” and “to.” Therefore a Relation establishes a directional link between two Nodes. A Node represents a concept, and a Relation denotes a relationship that exists between two concepts. The triplet (Node 1, Relation, Node 2) can be interpreted as a sentence wherein Node 1 is the subject, Relation is the verb, and Node 2 is the object—the classical ­subject/verb/object (SVO) structure used in natural language typology. By grouping multiple triplets, it is possible to define directed graphs of Nodes and Relations that describe complex relationships between concepts. The specialization of Nodes and Relations in their respective class hierarchies permits the capture of specialized meaning.

A semantic network can be created at runtime and therefore allows for dynamic semantic descriptions of notions that have not been considered beforehand. Multiple inheritance is addressed simply by linking a Node to multiple other Nodes through different relations. Through an example of this methodology detailed in the complete paper, the authors describe two signals that share the same physical quantities and yet have two different meanings, one being a physical property of a material while the other is a pressure gradient.

Networks of Logical Positions To Represent a Drilling System

It is possible to distinguish the nature of two different signals having the same physical quantities by constructing semantic networks using only a few numbers of Nodes and Relations. However, it is not possible to know from where those two signals originate. To solve that problem, the authors introduce another network, this one representing a logical description of the drilling system from a hydraulic perspective. By logical description, the authors mean a depiction that allows the capture of various logical functions of the drilling system. This is to be contrasted with a physical description that would have detailed every single physical element of the drilling system, even if some of those components would have had no direct effect on the logical behavior of the drilling system.

A semantic network description may focus on a subset of the hydraulic logical circuit. If the new semantic definition necessitates modification of the existing hydraulic logical circuit, then the modification is made within a transaction in order to avoid concurrent modifications leading to inconsistent results. If the result does not respect the cardinality of the inlets and outlets defined for each element in the hydraulic circuit, or if hydraulic state objects are lacking, then the transaction is rolled back; otherwise, the change is committed.

There are two other logical descriptions of the drilling system. One is the mechanical perspective, and the other one corresponds to the heat-transfer perspective. The reason for these multiple logical views is simply that some logical elements may exist in one perspective and not in others. For instance, the mud pumps, shakers, and pits do not play any role from the mechanical viewpoint of the drilling system but are essential elements from the hydraulic perspective. Similarly, most of the logical details of the hoisting system are irrelevant from the hydraulic perspective.

Semantic Network and Semantic Rules

A semantic network allows the expression of facts as SVO triplets. Therefore, each Node in the semantic graph can be both the subject of several SVO triplets and the objects of other SVO triplets. As with natural languages, it is unlikely that all possible combinations of SVO sentences, sharing the same word, make sense. For this reason, the authors supplement the semantic network with another level of semantics that focuses on allowed or necessary combinations of SVO triplets that start (i.e., the Node is taken as a subject) or end (i.e., the Node is considered as an object) from a specific Node. This semantic level is expressed as a set of rules.

Conclusion

  • A methodology to express the semantics of drilling signals has been described.
  • A semantic network that relies on exposing facts as SVO triplets defines a first level of semantic description.
  • This system is complemented by a set of semantic rules that ensures that the combination of different facts stays coherent.
  • The semantic description relies on a certain number of general principles.
  • It is not a static description and, therefore, allows for the characterization of a large variety of existing and future real-time drilling signals.
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 194110, “Toward Seamless Interoperability Between Real-Time Drilling Management and Control Applications,” by Eric Cayeux, SPE, Benoît Daireaux, Nejm Saadallah, and Sergey Alyaev, NORCE, prepared for the 2019 SPE/IADC Drilling Conference and Exhibition, The Hague, 5–7 March. The paper has not been peer reviewed.

Semantic Framework Aligns Real-Time Drilling-Management and Control Applications

01 February 2020

Volume: 72 | Issue: 2

No editorial available

ADVERTISEMENT


STAY CONNECTED

Don't miss out on the latest technology delivered to your email weekly.  Sign up for the JPT newsletter.  If you are not logged in, you will receive a confirmation email that you will need to click on to confirm you want to receive the newsletter.

 

ADVERTISEMENT

ADVERTISEMENT

ADVERTISEMENT

ADVERTISEMENT