As our clients have been dipping their toes into AI solutions, we’ve researched and evaluated a number of platforms to assist with further automation and advanced analytics given the overwhelming amount of data that operational teams work with. It’s difficult to stay on top of 50,000 tags and 200 screens daily, much less meaningfully analyze this data. AI works autonomously to call attention to what’s most important and prevent critical errors from slipping through the cracks, making it an invaluable timesaving tool for operators.
TwinThread (as well as AVEVA Advanced Analytics), has emerged as one of the most robust, with a large set of AI and machine learning (ML) features and an intuitive user interface. Their solutions can be applied across various sectors including water and wastewater, energy, and manufacturing.
The most valuable feature of TwinThread are its out-of-the-box AI and ML algorithms. The platform allows users to select an algorithm, such as predictive maintenance, and plug in their data to get a working model without needing to develop custom solutions. The platform features a built-in dashboard that provides operators with a visual representation of asset behavior trends. TwinThread can run workflows independently when specific conditions are met; such as detecting anomalies or errors. It sends alerts to operators and can issue commands back to SCADA systems to fine-tune or resolve issues without requiring operator intervention, thereby reducing labor hours.
TwinThread offers a range of additional capabilities in:
The result is a reduced workload with minimal error. TwinThread alleviates the burden of manually monitoring tags and assets while also maintaining a high standard of stability and consistency in system monitoring and management. TwinThread’s findings also make it easier to draw conclusions and identify areas for improvement in cost efficiency. By taking the guesswork out of analyzing tag data, teams can concentrate on addressing issues and implementing improvements.
While TwinThread’s algorithms may be low-code or no-code, the model does require training.
Training TwinThread follows a process similar to that of other AI tools. It begins with gathering data from various sources, including historical data from existing systems like SCADA and MES, as well as real-time data from sensors. High-quality data is crucial, so it must be cleaned to eliminate inaccuracies, duplicates, and irrelevant information. This preparation may also involve transforming the data into a suitable format for analysis. Once the data is cleaned and prepped, TwinThread automatically runs through the model training and testing cycles to produce a fully trained model.
This low-code environment greatly accelerates modeling workflows and democratizes this technology. However, systems can be very complex and difficult to model with out-of-the-box solutions. Our engineers provide expertise in assessing, customizing, and fine-tuning model performance. For complex modeling tasks, the first model produced may be underfitted, overfitted, or too noisy. These issues can be combated with custom cost functions, expanded training datasets (calculated or simulated), or through a variety of other methods.
When the model has been trained and validated, it can be deployed into the production environment. The performance should be closely monitored to ensure its effectiveness over time and may require adjustments. As the model collects new data and insights, this can enable periodic fine-tuning of the model to adapt to changing conditions.
Finally, as the model ingests real-time data, it can provide ongoing analysis and insights that inform your next steps. Once fully implemented, the system runs autonomously. On the user end, TwinThread’s digital twin technology creates virtual representations of physical assets and processes, allowing for precise monitoring and management. This involves modeling various parameters and signals associated with equipment, such as energy usage and operational metrics.
AI solutions should only be a priority for organizations that have reliable existing infrastructure for data collection and management. For industrial facilities, this means that the SCADA system, PLC programs, and data storage solutions should already be well-functioning before solving their next problem. It’s not only the initial model that requires a reliable collection of high-quality data. A consistent feed of clean and well-structured data helps TwinThread continue to stay accurate.
As a cloud-hosted platform, TwinThread has very light requirements for on-premises hardware. However, this means that organizations need to have strong cybersecurity measures in place. This includes having a segmented network and proper firewall configurations to secure data connections between on-premises systems and the cloud.
If you’d like to find out where AI solutions like TwinThread can fit in your technology stack, contact us. Our services include everything from model design to deployment, and we can provide the necessary services to ensure supporting infrastructure is ready for AI technology. Though we have vetted TwinThread, we are software agnostic and provide vendor selection services and training so your staff are equipped to work with the software best suited to their needs.