verticale

Manutenzione 4.0 e Oil & Gas: le nuove frontiere per la Manutenzione di Impianti

(in lingua inglese)

- Intelligent automatized production of deliverables
- Machine learns from human (supervised learning)
- Human learn from machine (e.g. unsupervised learning, decision tree etc.)
- Machine learns by making mistakes and by selection (genetic algorithm)
- Data intensive, algorithms are ready and strong, they just need good data.

Scarica il PDF Scarica il PDF
Aggiungi ai preferiti Aggiungi ai preferiti


Atti di convegni o presentazioni contenenti case history
mcT Petrolchimico Milano novembre 2017 Petrolchimico 4.0: tecnologie innovative per la progettazione e la conduzione degli impianti

Pubblicato
da Alessia De Giosa
mcT Petrolchimico Milano 2017Segui aziendaSegui




Settori: 


Estratto del testo
Milano, 30 novembre 2017 Gli atti dei convegni e pi di 8.000 contenuti su www.verticale.net mcT Petrolchimico 30 Novembre 2017 'Petrolchimico 4.0: tecnologie innovative per la progettazione e la conduzione degli impianti" Maintenance in Industry 4.0
The New Frontiers and Concept
in the Oil and Gas Crown Plaza ' San Donato Milanese
30th Novembre 2017
Josh Ferrara
Claudio Nespoli
Regina Meloni
Marco Scognamiglio
' Index
- XSIGHT DIGITALIZATION and services for Industry 4.0
- Predictive Maintenance (case study)
- Other Application on going: 3d printing 12 PROJECT DEFINITION EX P ER T IS E Field Development
Planning
Project Full Life-Cycle Concept Definition (pre-FEED) XSIGHT DIVISION - WHAT Selection an optimal development
concept for a Project.
Add a further level of detail for more
precise definition, a better cost
estimate and selection of a single
option for FEED.
Convert the FEED results to deliver an
engineering design which can be
safely and efficiently constructed,
installed and commissioned. Work
closely with vendors.
Develop studies to extend the life
cycle of the field. Develop studies for the decommissioning of existing facilities, providing the best solutions in terms of security,
cost, environmental impact, social
and risk management. ENHANC E Technology Licenses Project Definition
Basic Design - PDP
Front End Engineering Design
(FEED)
Engineering Services
Late Field Life Operations Brownfield
Modifications
De-bottlenecking Decomm., Abandonment, etc Master Plan Conceptual Design Technological Development Alternative Design Evaluation Process Optimization Hazop/Hazid/SIL Management Enviromental and Social Studies Technology licensing Technology Innovation Project Consultancy Screening of Project development
options.
Project Management Services Contracting Permitting Stakeholder Management Currently Digitalization within XSIGHT ' Opportunities ' INDUSTRY 4.0 ' Sustainable Engineering Solutions ' Sustainable Operation - Project Data and Information Management Solutions ' Digital Knowledge Implementation 14 Maximaze the application of the new insights innovation for the
O&G Industry to provide Capital Porojects and boost new Business:
' Big Data ' Advanced Analytics ' IoT ' Internet of Things ' Machine Learning ' Business Intelligence ' Cognitive Intelligence ' Collaborative Integrated Platform XSIGHT Business - The digitalization with the new technology leveraging on this new technology to enhance the level of completeness and of quality of the related services for potential Customers Master Plan Conceptual Design Technological Development Alternative Design Evaluation Process Optimization Hazop/Hazid/SIL Management Enviromental and Social Studies Technology licensing Technology Innovation Project Consultancy Project Management Services Contracting Permitting Stakeholder Management 1. Predictive Maintenance 2. Preventive maintenance optimization study 3. Plant Life Extension Study 4. Training & Advanced Automation Solutions 5. Wearable devices and AR/VR 6. Additive Manufacturing - 3D Printing Possible Services and Opportunities XSIGHT ' ADVANCED Maintenance Services ' Application with Neural Network
' Machine Learning ' For Spare PARTS
' For Design Optimization 16 INDUSTRY 4.0 New and current plants represent big opportunity for optimization through
innovative services based on the fol owing technologies: Technology and digital revolution is now impacting Oil&Gas industry 17 ' Reduces spare parts inventory costs
keep a couple of spares on hand to avoid machine downtime. perfect 3D printing fit solutions against the complexity of the design ' Reduces assembly costs ' Replace discontinued parts
the machine become outdated and spare parts may be difficult to find or too
expensive to order. Keep an older asset running longer by outsourcing the 3D
printing of discontinued, high-value parts. XSIGHT ' 3D Printing Simplified steps for replacing a part through 3D printing 3D printers with a variety of sizes and functionality, Scanners are mounted or hand-held, and use lasers or other types of light to capture the size and shape of an object. Materials, such as various plastics, metals, rubbers or even wood and glass, can be used to create objects. The materials used wil affect the durability and temperature resistance of the final product. ADDITIVE MANUFACTURING ' 3D Printing to Improve Maintenance Management - The critical items are selected by Criticality analysis (RAMS, FMECA) and RCM\RBI - Predictive Maintenance with cloud architecture with Secure Internet Connection - Hystorical data Predictive Maintenance Predictive Maintenance 20 Artificial intelligence applied ' Case Study Artificial Intelligence / Machine Learning 21 1) High Number of people involved
using collaborative platforms
2) Common target Enhanced rate of development 3) Continuous knowedge
build-up
' Google is leader in deep learning, moreover DeepMind is involved also in long-term research project Artificial Intelligence / Machine Learning 22 ' Predictive Analytics, example of applications: a) predictive maintenance
b) find correlation between an event of interest and historical sensors data ' Typical phases of the predictive analysis: 1) problem definition (business understanding)
2) selection of sensors (data understanding)
3) machine learning (data preparation and model ing)
4) root cause analysis Predictive Analytics ' Case Study Approach 23 Predictive Analytics ' Case Study Business Understanding 24 ' Brief definition of problem - Granulation scrubbers extraction fans abnormal vibration - Fan is equipped with no contact vibration probes connected to control room for a continuous shaft vibration monitoring. - Since the fan is working in dusty service it is necessary to stop the fan and to clean the internals in order to remove the urea stacked on the impeller and other components. This urea layer is normally the cause of the fan unbalance with the consequent increase of vibration. Predictive Analytics ' Case Study Business Understanding 25 Predictive Analytics ' Case Study Data Understanding 26 - Trend of data related selected process variables, which one correlates with abnormal vibration on extraction scrubber fan ' - 6 months historical data, 5 minute frequency of data value - Around 4000 samples (about 1 hour each) into training, test and validation samples Predictive Analytics ' Case Study Data Preparation 27 - Clustering time series unsupervised learning to group different time series data - PCA unsupervised learning to group different features - Feature selection model has to be as lean as possible, only more important features to take - Outlier removal noise and errors to be removed at the source - Control over data has to be done through time series visualization - Each relevant part of time series could be addressed by a different algorithm (boosting) Predictive Analytics ' Case Study Modelling 28 Mean Absolute Error on al test samples = 2.28 % Machine Learning ' Neural Network True and predicted value - test set ' 25 minutes ahead 29 Machine Learning ' Neural Network True and predicted value ' validation set ' 25 minutes ahead Vibration
(standardized)
Water Flow-Rate
(Standardized)
Warning 25 minutes before
occurrence
Vibration increase
occurring
30 - Machine Learning Analysis (e.g. neural network) al owed not only to identify among several variables a correlation between flow-rate of injection wash water and fan vibration, but also to identify its characteristic time (influenced by fan inertia). - Practical consequences: change of operating procedure, al owing operators to use that manual wash water valve only once the flow-rate has been duly reduced. Nevertheless, to avoid any impacts in case of operating mistake, the fan structure was reinforced in the direction of the flow (x axis). - Potential fan damage and/or loss of production avoided. Predictive Maintenance Conclusions 31 - Intel igent automatized production of deliverables - Machine learns from human (supervised learning) - Human learn from machine (e.g. unsupervised learning, decision tree etc.) - Machine learns by making mistakes and by selection (genetic algorithm) - Data intensive, algorithms are ready and strong, they just need good data. Artificial Intelligence for Design ' planned path forward 32 Any Question' Contact josh.ferrara@saipem.com regina.meloni@saipem.com claudio.nespoli@saipem.com marco.scognamiglio@saipem.com


© Eiom - All rights Reserved     P.IVA 00850640186