276660367 | Altitudevs | Dreamstime
Dreamstime M 276660367

How to harness operational data to meet sustainability targets

Nov. 8, 2023
There are many lessons to be learned from manufacturing companies that have improved their environmental impact—and their bottom line at the same time—by strategically placed sensors and correctly interpreting the readings they yield.

Manufacturing companies have a heavy burden to meet sustainability demands. In addition to individual process complexities, companies share concerns such as water scarcity, post-production wastewater treatment, energy conservation, and emissions reduction.

A recent study indicates that consumers and investors care about the environmental effects of making the products they use. Products from manufacturers who share their efforts to achieve environmental, social, and governance (ESG) goals averaged 28% cumulative growth in the past five years. Companies that made no such claims to meet sustainability goals saw only a 20% cumulative growth during the same period.

See also: How end-user computing in manufacturing is being revolutionized

Manufacturers looking to improve their environmental effect and public perception can find answers in operational data. Leveraging operational data promises a more agile and profitable future, but it also has the potential to improve health, safety, and the effects on the environment. Operational data lets manufacturers solve their unique challenges and achieve their bottom-line goals. In fact, companies have saved millions of dollars in a short period of time by optimizing their factories with the help of operational data.

Becoming a data-driven manufacturing operation

Industrial processes generate vast amounts of time-series data from sensors located throughout the factory. Many other operational systems generate data of their own as well. These include maintenance records, shift reports, and data from lab information management systems. This information, known as contextual data, helps put time-series data into context and provides engineers with a holistic view of manufacturing operations.

The information contained in operational data allows engineers to learn about the behavior of their manufacturing processes over time. By applying advanced industrial analytics, local experts are empowered to explore the data themselves to answer most of their everyday questions. When a more robust solution is needed—such as an anomaly detection model or classifier—engineers and data scientists can use advanced industrial analytics to collaborate on a machine learning exercise.

See also: Cybersecurity stakeholders praise AI executive order—but say it’s just a start

With the help of operational data, engineers can learn about raw materials usage, energy consumption, and overall equipment effectiveness. They can use this information to solve process anomalies and predict when maintenance will be needed. As operational performance improves, manufacturers get closer to achieving their sustainability goals.

In the chemical industry, for example, engineers can use advanced industrial analytics to compare information about batches. They can use this to determine what factors contribute to a better batch and save the ideal batch process as a golden fingerprint to be used in the future. In the food and beverage industry, engineers use operational data to preserve food quality and ensure its safety. This information can be used to prevent costly shutdowns of critical equipment.

Reducing batch cycle times using advanced analytics

In one example, a chemical plant produces specialized fibers. These fibers, while essential for coatings and films, also have applications in the pharmaceutical industry. This requires stringent quality controls. The production lifecycle involves a multistage batch reactor process, with each stage transition being meticulously managed. However, over time, inefficiencies evolved in the reactor. Production cycles began to get longer, yields declined, and efforts to solve the problem were unsuccessful.

Using advanced industrial analytics, operational experts looked at the reactor’s behavior over time. They could see that midway through the full batch cycle, recent batches deviated significantly from the best performing batches of the past. Engineers immediately realized this was related to a change made in a raw material feeder. They were able to adjust the feeder controller and map it to historical best performance. Just four hours after engineers began their investigation, the improvement resulted in a 30-minute savings on a nine-hour batch cycle. This translates to $1 million per year in savings.

Increasing energy production after reviewing operational data

In another situation, a plant in the energy and utilities industry uses natural gas to produce electricity. This combined-cycle gas power plant has two turbines that run a generator. Over time, the performance of a unit of the plant’s power station began to decline, which led to capacity and revenue loss. The loss was gradual, so it went unnoticed at first. After a couple of years, however, it became clear that performance was degrading. The team wanted to quantify the impact and duration of their losses.

See also: Why manufacturers' data is best analyzed at the edge

After reviewing operational data and filtering out irrelevant periods, the average power production was shown to have decreased about 2% over the course of three years. With the help of advanced industrial analytics, operational experts were able to see that the gas fuel flow, the compressor discharge temperature, and the inlet guide vanes reference angle differed consistently across two different groups over time. After further investigation, they concluded that the performance issue was the result of non-calibrated inlet guides. Solving the airflow anomaly led to an additional 2 megawatts of power production and $260,000 in value.

Applying an ML technique known as anomaly detection

Some anomalies are much more difficult to detect and predict. In certain situations, an anomaly might happen even when a process appears to be functioning normally. For these more involved cases, data scientists can apply a machine learning (ML) technique known as an anomaly detection model.

In one case, fluid from one process occasionally leaked into a compressor at a chemical plant and damaged it. This led to a full-plant shutdown and significant loss of production. Engineers already had determined that vibrations were causing the leak, but they had no way to monitor this problem. Operational experts could not intervene in time to prevent damage and avoid the shutdown.

See also: Real-time simulation's role in driving business competitiveness

Engineers attached vibration sensors to the compressor. They then used advanced industrial analytics to collaborate with data scientists on creating the model. Once operational data was used to train the model with different types of vibrations, they were able to create an ML tag that acts as a monitor for irregular vibrations. When the anomaly was detected, operational experts received an alert with enough time to make changes before the compressor was damaged and a complete shutdown occurred.

Sustainability demands are not going away. As the world looks for companies to improve their environmental effect, manufacturers must find ways to meet the needs. By becoming data-driven, operational experts find ways to optimize operational performance, improve production, increase their bottom line, and leave behind a healthier planet.

About the Author

Wouter Daniels

Wouter Daniels helps train and advise TrendMiner customers to get the most out of their operational data. He also provides custom solutions for companies with more complex cases.