Aspen Technology : The Aspen Industrial AI Workbench

In May 2021, Aspen Technology announced that it is further extending Industrial AI across its leading solutions to drive higher levels of profitability and sustainability in industrial operations. As part of our V12.1 release, the Aspen Industrial AI Workbench™ was brought to market, enabling data scientists to collaborate with domain experts to develop AI apps based on enterprise-wide data. Enterprises are looking for purpose-built, fully integrated AI environments for their data scientists to accelerate the transformation from raw data to productized AI/ML (machine learning) algorithms, working hand-in-hand with subject matter experts.

The Aspen Industrial AI Workbench™ is part of the Aspen AIoT Hub™ - it includes both the Aspen IoT Analytics Suite™ and Aspen Data Science Studio™, combining out-of-the box analytics libraries and a production-grade AI collaboration environment to empower both domain experts and data scientists.

Aspen Data Science Studio
Aspen Data Science Studio provides an AI development environment for data scientists, with an embedded workbench for feature engineering, training, collaboration, versioning and the rapid productizing of ML models - all in a user-friendly self-service interface without the need for any complex setup.

Aspen IoT Analytics Suite
To help bridge the gap between data scientists and domain experts, Aspen Industrial AI Workbench also includes Aspen IoT Analytics Suite. It's a set of ready-to-use analytics tools built with industrial managers, data analysts and domain experts in mind - with a minimal configuration and no-coding interface to explore industrial data, generate prediction, visualize KPIs and share reports and dashboards with other stakeholders.

Empowering data scientists in the industrial world

When building and collaborating on Industrial AI and ML models, there are several questions that data scientists ask themselves:

• How to connect to a specific data source or multiple data sources? In particular, access to real-time and historian data is often difficult and requires IT involvement
• Where to put the ML code? How to version it? How to make sure you're using the right set of libraries?
• What if one needs to collaborate with domain experts as well as enterprise users? Data scientists often need to incorporate domain knowledge which requires tedious back and forth with subject matter experts, many scattered files and disparate tools
• What if data scientists need to share their results and deploy their model in production? While developing PoCs is often straightforward, most PoCs don't get productized because the effort and underlying infrastructure involved often requires the help of IT, Data Engineers, DevOps experts and integrations efforts.
• …and, then what if one needs more computing resources to run their ML models based on the increased complexity or increased amount of data? There is a clear need for a ready-to-use, industry-grade, full-stack solution that is fast, scalable, includes monitoring as well as tight security parameters.

Through the Aspen Industrial AI Workbench application data scientists don't have to figure this all out. The application is hosted on a cloud-ready and built-for-industry Industrial AI infrastructure solution, the Aspen AIoT Hub. The premise of the Industrial AI Workbench is to let data scientists focus on what they do best: develop domain-specific ML models. The solution abstracts all data infrastructure complexity involved in setting up an industrial data science project, streamlines the path to ML productization and makes it easy to fetch the required data. Furthermore, it provides an environment to collaborate with domain experts such as process engineers and deploy projects in production seamlessly, at the click of a button.

Key benefits and capabilities

• Enhanced collaboration on Industrial AI models: Bridges the gap between data scientists and subject matter experts via a common interface. This enhances overall team productivity, flexibility and collaboration by developing and running code from anywhere with hosted Jupyter notebooks. The product also streamlines the development of IP and integration of domain expertise by importing custom dependency sets. Users can improve future model iterations with close tracking of code changes and track quality output of ML models.
• Simplified infrastructure, integration and deployment: Easy adoption and integration into business workflows with a turnkey, API-driven and cloud-agnostic solution. The product is hosted on the AIoT Hub which provides the integrated data management, edge and cloud infrastructure and production-grade AI environment to build, deploy and host Industrial AI applications at enterprise speed and scale. The application leverages the AIoT Hub ecosystem with existing integrations to Aspen InfoPlus.21™ Data Historian, Aspen Connect™ and Aspen Enterprise Insights™. Furthermore, it delivers optimal performance during ML training and testing by leveraging highly efficient Spark clusters.
• Empower enterprise users with advanced IoT analytics libraries: High-value insights through native out-of-the-box IoT libraries which addresses analytics needs of domain experts and enterprise users without technical or coding skills. This enables domain experts with data analytics capabilities via a no-code visual query builder and empowers citizen data scientists with advanced analytics libraries such as anomaly detection, predictions, scoring and many more - thereby accelerating time-to-insight.

Getting ahead in your Industrial AI strategy

Launching new Industrial AI projects and managing the full lifecycle of AI/ML presents a number of challenges, and very often they are not of the nature and scale initially planned. But while the development of AI/ML is inherently challenging (e.g. training sets, availability of data scientists, domain expertise, etc.) from a product point of view, the actual AI/ML code makes up a very small portion of the overall solution.

In fact, when it comes to industry-grade AI projects, deploying and running AI/ML code against live streaming data, as well as maintaining it over time, requires a specific level of expertise that goes well beyond the original challenges of AI/ML. Add to that the need for live collaboration between domain experts and data scientists, the growing need for enterprise-wide data insights, and the integration of one's intellectual property (IP)…. and you have yourself a very complex and expensive puzzle to solve.

It is with the goal of alleviating this complexity in mind that AspenTech has developed and launched its Aspen Industrial AI Workbench application. This fully-managed, industrial-focused AI/ML environment allows customers to develop, train/maintain, deploy, and run Industrial AI/ML models as well as ad-hoc algorithms against live streaming data - underpinned by a robust, highly-scalable and secure Industrial AI infrastructure.

To learn more about the Aspen AIoT Hub, visit our solutions page.