Documente Academic
Documente Profesional
Documente Cultură
net/publication/299397674
CITATION READS
1 831
5 authors, including:
Mario Hoernicke
ABB AG Corporate Research Germany
87 PUBLICATIONS 145 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mario Hoernicke on 24 March 2016.
ESTEBAN ARROYO, STEVE ROYSTON, ALEXANDER FAY, accessible and up-to-date format can require a gargan-
MARIO HOERNICKE, PABLO RODRIGUEZ – Documentation in tuan effort. ABB has explored means of efficiently extract-
process industries can rise to bewildering dimensions. ing models from engineering documents in an automated
Furthermore, the difficulties of accessing it can be and consistent manner. A joint research project by ABB
compounded by the range in formats used, including and the Helmut Schmidt University (HSU) in Hamburg,
traditional paper and elementary digital representations. Germany, resulted in a method based on optical recogni-
When a plant owner decides to consolidate this informa- tion and semantic analysis that converts piping and
tion – for example, to prepare a modernization process instrumentation diagrams (P&IDs) into object-oriented
– bringing this eclectic documentation into a single, models.
L101 L101
Pi001 Pi002
V200
E100
T201 T201
V300
F102 F102
Pi003 Pi004
tion with HSU, devised methods for the cy papers or CAD files ➔ 3. From here, the
extraction of
object-oriented (OO) repre- method executes three procedures for the
sentations from existing design docu- generation of the computer-interpretable
ments, specifically P&IDs [2]. model, namely:
– Optical recognition: OSR and
Previous research has shown that the use OCR ➔ 4 for the identification of
of P&IDs can support different tasks along symbols and text annotations.
the plant’s life-cycle such as automated – Semantic analysis: Interpretation of
HAZOP (hazard and operability) studies [3], the functional connotation of graphi-
detection of design patterns [4], derivation cal forms for the enhancement of the
of simulation models [5] and fault detection recognition procedure and the
T
and diagnosis [6]. Existing approaches are improvement of expression capabili-
he sheer size and complexity of all based on the prerequisite that P&IDs ties for the model description.
process-industry plants com- must be described
bined with the numerous modi- in OO formats (eg,
fications and additions effected IEC 62424 CAEX/ Input documents are com-
throughout their life-cycles results in vast AutomationML [7,8]
amounts of legacy documentation. This or ISO 15926 [9]). posed of four fundamental
includes structural models, functional
descriptions, P&IDs and equipment spec-
Unfortunately, this is
not typically the case
artifacts, namely parametric
ifications. Although these documents in existing process forms, nonparamertric forms,
embody a rich source of information that facilities as most
should be exploited in planning and op- P&IDs are filed in text annotations, and under-
erational activities, this is currently hin-
dered by the challenges entailed by their
non-computer-inter-
pretable forms.
lying semantic content.
timely localization and data extraction [1].
From pixels to meaning – Representation and verification:
In current industrial practice, existing engi- The method extracts OO models from Description of captured information as
neering documentation is digitized manu- mixed text-graphics documents, con- an OO model followed by visual
ally by scanning, storing and indexing. The cretely P&IDs ➔ 1, recognizing text, sym- inspection.
resulting files contain unstructured infor- bolic forms and connection topologies.
mation that cannot be fully exploited for The approach relies on the fact that input Optical recognition (OR)
automation tasks as their underlying con- documents are composed of four funda- OR deploys sound image processing tech-
tent does not take the form of a data mod- mental artifacts, namely parametric forms, niques for the identification of geometric
el. Addressing this, ABB has, in coopera- nonparametric forms, text annotations and forms and text identifiers. The process is
underlying semantic content ➔ 2. executed through the recursive application
of two methods: OSR and OCR.
The conversion process assumes that
P&IDs are stored in pdf files (pdf/bitmap or Optical symbol recognition
pdf/SVG) from which raster images can be As the first step of OSR, nonparametric
Title picture
extracted. In some cases, however, pre- symbols (ie, multi-curve forms) such as
Documentation in process industries can take
various formats, including paper and primitive
liminary process steps are required so that “vessel E100” and “valve V300” shown
electronic formats. How can all this be translated a raster image can be subsequently ob- in ➔ 1 are localized and matched against
to a modern object-oriented standard? tained, particularly when dealing with lega- predefined libraries of templates. A library
P&I
diagram PDF/bitmap Raster
PDF/SVG
image
Optical recognition
Proposed method
Circle Line
Virtual printer
Scanner
Semantic analysis
Curve
CAD
Paper
Representation & verification
Semantic content
can represent, for instance, plant asset and thereby avoid the misinterpretation of
4 Glossary of terms
catalogs or sets of symbols commonly characters as symbols. As with the non-
used to embody specific devices or pro- parametric symbols, the reliability of the AML AutomationML
cesses within P&IDs. The definition of recognition process can be adjusted CAD Computer-aided design
of the examined forms. Three key geomet- Based on the notion that the text identifier OR Optical recognition
ric properties ➔ 5 are supported to guar- of an object is located in its surroundings OSR Optical symbol recognition
antee an exhaustive search of possible or within the object, the coordinates of SA Semantic analysis
matches – namely, rotation, scaling and identified forms are used to generate
SVG Scalable vector graphics
occlusion. search windows ➔ 6. The size of a window
is typically proportional to the dimensions
Scores quantifying the extent of resem- of the analyzed image. It can also be de-
blance between identified candidates and fined by the user. An OCR algorithm
templates are used to define levels of con- searches for text within the defined win-
dow. If more than
one tag is found,
A semantic rule might state, nomenclature rules
(eg, regular expres-
for instance, that a pipeline sions) are used to
cally corrected
403 403
F F
Pi157
V112
J105 I
HS 105 I TE 116
I
I
Search window
I: Information flow connector
E100 (vessel)
E100-L101 (InformationFlowConnector
E100-Pi101 (MaterialFlowConnector)
- ID
- Tag name
Pi001 (pipe) IL - SizeY
Pi001-E100 (MaterialFlowConnector) - SizeX
- PIDYcoordinate
- PIDXcoordinate
L101 (sensor) IL - Height
- Diameter
L101-E100 (InformationFlowConnector)
OO Modeling References
An OO model, specifically a CAEX/Auto- [1] E. Arroyo et al., “Integrating Plant and Process
Information as a Basis for Automated Plant
mationML model compliant with IEC 62424
Diagnosis Tasks,” IEEE International Conference
[7,8], is derived from the collected data. on Emerging Technologies and Factory
The procedure starts by creating an object Automation (ETFA), Barcelona, Spain, 2014.
for every entry within the first column of the [2] Diagrams for the chemical and petrochemical
industry, DIN EN ISO 10628:2001.
connectivity matrix. Subsequently, inter-
[3] A. Fay et al., “Knowledge-based support of
faces are added. For example, for vessel HAZOP studies using a CAEX plant model,”
E100, two interfaces are created – one of Inside Functional Safety, vol. 2, pp. 5–15, 2009.
type MaterialFlow (F) and one of type Infor- [4] R. Drath et al., “Computer-aided design and
implementation of interlocking control code,” in
mationFlow (I) ➔ 10. The process con-
IEEE International Symposium on Computer
cludes by assigning attributes (eg, position Esteban Arroyo
Aided Control Systems Design (CACSD), 2006,
and dimensions). Thus an OO model de- Alexander Fay pp. 2653–2658.
scribing the connectivity and semantic Helmut Schmidt University [5] M. Barth et al., “Object-oriented engineering
data exchange as a base for automatic
information of the P&ID is obtained. Hamburg, Germany
generation of simulation models,” IEEE
esteban.arroyo@hsu-hh.de
Industrial Electronics Society (IECON), Porto,
Visual inspection alexander.fay@hsu-hh.de Portugal, 2009.
Finally, a graphical depiction of the inferred [6] S.Y. Yim et al., “Using process topology in
plant-wide control loop performance assess-
information is generated ➔ 11. The repre- Mario Hoernicke
ment,” Computers and Chemical Engineering,
sentation overlays the original black and Pablo Rodríguez
vol. 31, pp. 86–99, 2006.
white P&ID with colored forms represent- ABB Corporate Research [7] Representation of process control engineering
ing the identified objects. Users can visual- Ladenburg, Germany – Requests in P&I diagrams and data exchange
ize these objects and compare them with mario.hoernicke@de.abb.com between P&ID tools and PCE-CAE tools,
IEC 62424 (2008-08).
the original diagram, effecting changes pablo.rodriguez@de.abb.com
[8] R. Drath, Datenaustausch in der Anlagenpla-
where necessary. Corrections carried out nung mit AutomationML, Germany: Springer
in the graphical interface have a direct Steve Royston Verlag Berlin Heidelberg, 2010.
effect on the OO model, thus avoiding the ABB Process Automation, [9] T. Holm et al., “ISO 15926 vs. IEC 62424
– Comparison of Plant Structure Modeling
need for additional engineering tools. Pre- Oil, Gas & Chemicals
Concepts,” IEEE International Conference on
viously generated warnings can be used to St. Neots, United Kingdom Emerging Technologies and Factory Automation
prioritize the cross-checking procedure. steve.royston@gb.abb.com (ETFA), Kraków, Poland, 2012.