Leveraging the power of AI to improve profitability while enhancing asset reliability and safety

By Stephen Reynolds, Industry Principal for Chemicals, AVEVA

It’s safe to say that running a chemicals plant has never been more nuanced or challenging.

Today’s companies face myriad pressures from all directions, including increased competition and demand, safety concerns, and sustainability initiatives.

Overcoming these challenges requires plants to optimize assets and processes, maximize production, reduce unplanned downtime, and ensure reliable and safe operations that meet wider business goals.

Of course, technology is the key to improved performance. But while many companies have already adopted data infrastructure and analytics systems, today’s growing plant complexities call for even greater technology powers in the form of artificial intelligence (AI).

Forward-looking chemical companies are choosing to layer rigorous models and AI-based advanced analytics on top of their data historians, to extract value and gain faster and better insights into processes and assets.

Advanced process simulation is key

Next-generation process simulation can now calculate key performance indicators (KPIs) beyond those that can be measured directly in the plant, while predictive analytics can detect performance anomalies and predict asset failure. Taken together, these high-level insights enable operations teams to deftly analyze risk and devise plans that maximize efficiency and profitability.

As such, the implementation of AI solutions allows companies to proactively manage asset performance, minimize unplanned downtime and reduce maintenance costs.
Every chemical plant is on its own unique data-driven journey. While some operations are using data collection and analytics as a historian, others have moved further along the utilization curve and are using streaming calculations or advanced analytics.

Regardless of where a plant is in its journey, users can still find new ways to optimize plant operations to increasing profitability, while meeting safety requirements and meeting sustainability goals.

Moving from reactive to predictive maintenance

Chemicals companies around the world are choosing to leverage AI and Machine Learning (ML) models as deep learning tools to forecast an asset’s remaining useful life, giving teams critical information and prescriptive insights to analyze cost-versus-risk and devise plans that maximize efficiency and profitability.

Users can define leading indicators based on sensor and other operations data and use this information to detect even subtle changes in asset performance.

Once teams have identified an anomaly, they can use advanced AI tools to predict performance degradation and component failures and then work together to prioritize maintenance needs based on urgency, schedules, available teams, resources and spare part availability.

In addition to preventing asset failure, predictive AI-based guidance allows companies to minimize energy usage and compare asset performance, helping chemicals plants meet regulatory and contractual obligations.

With future insights into asset performance at hand, companies can take action to minimize inefficiencies that affect financials, gauge future consequences, assess risk, avoid disruption, and even increase customer satisfaction.

Enhancing Operations Digital Twin using process simulation

While process simulation and predictive analytics are well-established approaches, these tools have continually improved over the past few years. Now, real time data combined with first-principal process models can be used to create operations digital twin. This operations digital twin, enhanced with AI capabilities, can be used to identify most optimal operating conditions and act as advisors to operators to get more out their asset.

These operations digital twin gives engineers and technicians insight into unmeasurable process variables, allowing tools to proactively predict best operating conditions, to increase yield, reduce energy as well as reliability issues for rotating and stationary assets across the enterprise.

Equipment monitoring and maintenance programs

Traditional equipment monitoring programs rely on data measured throughout the process to inform maintenance decisions. For example, temperature and vibration data may be used to predict a variety of failure modes for a centrifugal pump.

By using historic data, reliability engineers can determine a baseline value for each measurement and configure alerts when values fall outside of this range. This is known as condition-based monitoring and is a simple way to begin using measured data to improve process reliability.

While condition-based monitoring is useful for assets with relatively stable operation, accounting for different operating windows or process modes can quickly become a challenge. Engineers may need to frequently adjust operating windows or deal with nuisance alarms that can quickly breakdown the efficiency and effectiveness of a predictive maintenance program. Instead of condition-based monitoring, many chemical companies now rely on predictive analytics as part of a robust Asset Performance Management program.

Predictive analytics combines real-time process data and AI to better predict equipment failures and provide Remaining Useful Life (RUL) estimates. This information can help reliability engineers plan what equipment requires maintenance and when it should be scheduled. Like condition-based monitoring, it relies on process measurements to establish operating baselines.

However, since predictive analytics solutions can leverage ML for multivariate analysis instead of simple condition-based alerts, it provides a greater range of detection across many operating modes compared to condition-based monitoring alone.

The ML models used in a typical predictive analytics solution can also help identify specific failure modes based on previous fault signatures for each piece of equipment.

By using a powerful combination of real-time data, predictive analytics, and first-principles simulation companies can improve asset reliability and reduce downtime by proactively predicting reliability and integrity faults for critical rotating and stationary assets.

In conclusion

Amid an increasingly cut-throat global chemicals industry, where pressures are mounting and balance between business growth and meeting decarbonization goal is critical, AI will be the key to optimizing forward-looking businesses through the intelligent streamlining of their operations.

By layering advanced AI on top of existing data infrastructure, chemical companies can take advantage of real-time and historical operations data to gain access to deeper, faster, and more valuable insights to support business objectives and the overall business strategy. This allows chemical companies to find the balance between risk-based and reliability-centered maintenance, improve overall performance, and avoid potential equipment failure.

With advanced insights at their fingertips, chemicals plant teams can make better, faster decisions that continually improve plant performance, safety, and sustainability—while increasing profitability.