Creating a ‘risk-radar’ to monitor hidden risks

Continuously identify hidden risks such as schedule delays, poor quality, budget over-runs, and resource bottlenecks. New AI, and Machine Learning tools allow companies to leverage data from past projects to make proactive, informed project decisions by continuously monitoring “risk radar”.

It all begins with verifying the problems that you are trying to solve:

• What is the risk of rework due to incorrect information?
• What hidden risks do we need to prepare and plan for?
• What activities are likely to be delayed?
• How likely is the project going to be delayed and by how much?

Potential benefits:

• Moving humans away from harm and hazardous environments.
• Intelligent infrastructure – supports efficiency and repeatability of traditional asset examination methods leading to reduced cost, improved productivity, better understanding of asset degradation.
• Digital delivery supports move to predictive maintenance regime & decision making.

Use AI to plot risks and how they develop over time

A connected data set helps provide a more accurate and complete view of the project by identifying ‘hidden’ data relationships and patterns/trends. You then use out-of-the-box benchmarks to compare current projects with projects previously delivered. These prebuilt models allow you to get started quickly without having to write any code. You create your own custom groups to ensure that comparisons are like for like, based on how you run your business to drive better prediction accuracy.

Common use cases

• Forecast any time series metric – product demand, revenue, # of service requests.
• Builds multiple models and automatically select the most accurate one of your businesses.
• Deliver forecasts with ability to explain, which brings transparency to predictable results.

CASE STUDY: Use of Artificial Intelligence Software Service to automate Peer Check Review process of Nuclear Operating Procedures in a Nuclear New Build (NNB).

Straight-Through Processing to update the Enterprise Asset Management (EAM) system in SAP Plant Maintenance (PM) without double entry. We demonstrated how data would be extracted from existing New Nuclear Build (NNB) procedures and other data (such as the existing NNB Equipment Database, eSOMS, Alarm Point Data, etc.

Challenges

• Developing the Parent/Child relationships (P/C) within PROMS dB for the Normal Operating Procedures (NOPs), Surveillance Test Procedures (STs), Periodic Test Procedures (PTs) and Local Alarm Response Procedures (ARPs)
• Generated a RO Database to generate Primary/Replica procedures.
• Reporting process blockages can be alerted on NEC4, HSEQ and commercial compliance.

Solution

• Converting and migrating procedures and other data (such as the existing Nawah Equipment Database, eSOMS, Alarm Point Data, etc.) via OCR digitisation from a static format (PDF, csv) into a SQL dB (PROMS).
• Creating Dynamic Procedures/Smart AI Templates via SharePoint Forms, ABBYY FlexiCapture, MS Power Apps and Power Automate

Value

• Use of Artificial Intelligence Software Service to automate Peer Check Review process, which ensured that their Writer’s Guides have been appropriately implemented.
• Use Natural Language Processing (NLP) to examine any Note, Caution and/or Warning to ensure that compliance to support and enhance the performance of procedure writers.

Automated Asset Inspections with Machine Learning

We can simplify anomaly detection models to automatically flags critical incidents. Machine Learning (ML) powered automated data analysis tool for general inspection that enables predictive maintenance by providing consistent and reproducible results.

Automatic identification of rare items, events, or observations in data

Anomaly Detection is the identification of rare items, events, or observations in data that differ significantly from the expectation. This can be used for several scenarios like asset monitoring, maintenance, and prognostic surveillance in industries such as utility, aviation, and manufacturing.

The Anomaly Detection Service will create customized Machine Learning models, by taking the data uploaded by users, using MSET algorithm, which is a multivariate anomaly detection algorithm to train the model, and deploying the model into the cloud environment to be ready for detection. Users can then send new data to the detection endpoints to get the detected anomaly results.

How can this benefit SPACE EAST Members?

POTENTIAL SOLUTION: Creating a ‘unified’ AI/ML platform spanning cloud services and data assets between Delivery Partners

Obtain analytics visibility and gain insights through artificial intelligence (AI) and machine learning (ML). These ML powered automated data analysis tools allow general inspection that enables predictive maintenance. Companies can then configure QC validation process and data alerts to monitor sensor data in case of exceedance  of safety thresholds. Afterwards, companies can use correlation alerts for defining alert thresholds dynamically in the  form of linear and quadratic relationships between two sensor metrics.

DATA ACQUISITION & MANAGEMENT

• Use FTP to automate file data ingestion.
• Use revision management to save different sensor configurations.

TIME-SERIES DATA ANALYSIS

• Utilise time series charts with trending, annotation, and analytic capabilities.
• Develop X/Y and scatter plots.

DIGITAL TWIN VISUALISATION

• Utilise 4D navigation to display historical versions of the models and IoT.
• Configure data alerts to monitor sensor data in case of exceedance of safety thresholds.
• Display map layers from ArcGIS, WMS and WMTS servers on 3D terrain.
• Visualise 3D point cloud captured with laser scanners.