Solution Brief
Prevent productivity and revenue losses due to asset anomalies
Predict issues ahead of potential failures with AI and machine learning
The issue
Asset-intensive industries with continuous operations, such as manufacturing, energy and transportation, are adversely affected by productivity and revenue loss if asset anomalies are not detected and fixed proactively. To be proactive, these organizations need a robust anomaly detection system driven by AI and machine learning (ML) that can predict issues and alert maintenance teams ahead of potential failures. However, operationalizing an analytical system that is adaptable, affordable and scalable can be challenging.
The challenge
Complex IoT data
Inaccurate, inconsistent or incomplete sensor data has insufficient quality for model training.
Challenging integration scenarios
It’s difficult to integrate with operational technology (OT) systems, such as the AVEVA PI System.
Limited data science capabilities
There’s a lack of resources with data science experience and the analytical standards needed to create and operationalize the right models.
Deployment challenges
Issues with model scalability and poor visibility into model performance exist.
High total cost of ownership
Costs are high due to a large analytics software footprint that includes data collection, storage, model creation and operationalization.
Lengthy time-to-business value
It takes longer to go from data to deployment, which delays realizing the benefits the business is truly gaining.
(Improve AI/ML deployment success rates; Implement a unified decisioning platform for automation and governance; Integrate, deploy and manage analytical decisioning rules)
Our approach
Our industrial intelligent asset monitoring solution helps subject-matter experts (such as plant operators and maintenance engineers) proactively detect asset anomalies that could lead to machine failures, unplanned downtime and safety risks. Users with in-depth knowledge of processes and assets don’t need advanced data science capabilities to rapidly create and deploy thousands of unique models.
We approach the problem by providing software and services to help you:
Override business rules with ML models
Replace cumbersome, outdated business rules with ML models that compare actual behaviors with expected behaviors to identify potential issues.
Improve operational efficiency
Identify anomalies in assets before they fail, improving both asset uptime and operational efficiency.
Employ analytics at scale
Train and deploy thousands of advanced analytical models using automated ML techniques.
Optimize alarms
Receive focused alerts that provide detailed information that can immediately be acted upon.
Reduce analytical software costs
Operate your modeling environment on demand, which can significantly reduce overall software costs.
SAS® difference
With guided workflows and automated machine learning capabilities, our solution covers the full spectrum of the analytics life cycle. Connect to different IoT data source systems, import historical data, remove outliers, create derived variables, configure model definitions, prepare training data, and build and deploy high-performing models to monitor processes and assets in real time.
Speed
- Use streamlined guided workflows to connect and clean data, plus define and deploy models seamlessly.
Precision
- Train thousands of models simultaneously, each tailored to the unique data of individual assets to ensure precise and accurate results.
Efficiency
- Facilitate teamwork through automated machine learning (AutoML) that enables diverse skills to collaborate.
Scalability
- Deploy thousands of models into your operational environment with a single click.
No-code environment
- Implement a real-time anomaly detection project without any coding required.
Case Study: Streaming Analytics for a Smart Building
Situation
A highly interconnected building system with hundreds of assets and thousands of sensors encountered performance issues and high energy consumption, leading to high maintenance costs. Not taking any action would have resulted in critical asset failures and even higher repair costs.
SAS Solution
- Examined the normal behavior of each asset in the context of the connected system as a whole and used automated machine learning techniques to do a real-time comparison of actual vs. expected behavior.
- Identified anomalies in critical sensor parameters of assets (e.g., anomalies in the fan speed of the air handler unit) and alerted users before an actual failure occurred.
Results
- Detected asset degradation patterns based on actual vs. predicted behavior.
- Prevented critical asset failures by enabling informed decisions that generated a high return on investment.
- Realized significant energy savings due to asset repairs occurring before any performance degradation.
Learn more at sas.com/iotsolutions