Data foundations in chemicals
A production planner in a chemicals plant uses AsscherAi to investigate yield drift by querying the batch log and the maintenance schedule, then compares the results with the recipe and feedstock data to identify the root cause of the issue and determine the best course of action to take.
A production planner in a chemicals plant checks the yield of a batch of polyethylene, only to find it is lower than expected, and wonders if the issue lies with the feedstock, the recipe, or the plant's equipment, and how this will affect the answers they get from AsscherAi.
Chemicals industry challenges
The chemicals industry presents unique challenges for enterprise data readiness for AI, particularly in continuous processes with tight yield and safety envelopes. Yield drift, where the output of a process changes over time, can be caused by a variety of factors, including feedstock variability, equipment wear and tear, and changes in operating conditions. This makes it difficult to establish a baseline for what constitutes "normal" performance, and can lead to incorrect conclusions being drawn from the data.
Plant-to-plant recipe differences are another challenge, as the same process can be operated differently at different sites, making it hard to compare results and identify best practices. For example, a recipe for producing polyethylene may be optimized for one plant's equipment and operating conditions, but not for another's. This can lead to inconsistent results and make it difficult to develop a comprehensive understanding of the process.
These challenges are not insurmountable, but they do require careful consideration and planning to overcome. By understanding the complexities of the chemicals industry and the challenges it presents, organizations can take the first steps towards developing a robust enterprise data readiness strategy for AI.
Field-level meaning and documentation
One of the key issues in enterprise data readiness for AI is the lack of clear field-level meaning and documentation. Column names in a database are often not sufficient to convey the meaning and context of the data, and can lead to misunderstandings and misinterpretations. For example, a column named "temperature" could refer to the temperature of the reactor, the temperature of the feedstock, or the temperature of the product, without further context.
This is particularly problematic in the chemicals industry, where small changes in temperature or pressure can have significant effects on the process. By providing clear and concise documentation of the data, organizations can ensure that the meaning and context of the data are understood, and that the right conclusions are drawn from it. This can be achieved through the use of data dictionaries, data governance policies, and other tools.
Systems and scope
Another challenge in enterprise data readiness for AI is the tendency to leave certain systems or datasets out of scope. This can lead to partial answers and a lack of comprehensive understanding of the process. For example, a model may be trained on data from the production process, but not on data from the maintenance or quality control systems. This can lead to a lack of understanding of how these systems interact and affect the overall process.
By including all relevant systems and datasets in the scope of the project, organizations can develop a more comprehensive understanding of the process and make more informed decisions. This can be achieved through the use of data integration tools and other technologies that enable the combination of data from multiple sources.
Gaps and inconsistencies
Stockouts, backfills, and other gaps in the data can also teach a model the wrong lesson. For example, if a model is trained on data that includes stockouts, it may learn to predict stockouts as a normal part of the process, rather than recognizing them as anomalies. This can lead to incorrect conclusions and a lack of understanding of the underlying causes of the stockouts.
By identifying and addressing these gaps and inconsistencies, organizations can develop more accurate and comprehensive models of the process. This can be achieved through the use of data quality tools and other technologies that enable the detection and correction of errors and inconsistencies in the data.
Limitations and alternatives
While AsscherAi can provide valuable insights and answers to questions about an organization's data, it is not a substitute for human judgment and expertise. There are also cases where AsscherAi may not be the best choice, such as when the data is highly unstructured or when the questions being asked are highly subjective. In these cases, alternative approaches such as manual analysis or the use of other AI tools may be more appropriate.
By understanding the limitations and alternatives to AsscherAi, organizations can make more informed decisions about how to use the tool and how to address their data analysis needs. For more information on how to get the most out of AsscherAi, please contact us.
Frequently asked questions
What are the unique challenges of enterprise data readiness for AI in the chemicals industry?
Yield drift and plant-to-plant recipe differences make it difficult to establish a baseline for normal performance and compare results.
How can unclear field-level meaning and documentation affect AI outcomes?
Unclear column names can lead to misunderstandings and misinterpretations of the data, causing incorrect conclusions to be drawn.
Why is it important to include all relevant systems and datasets in the scope of an AI project?
Leaving out certain systems or datasets can lead to partial answers and a lack of comprehensive understanding of the process.
What can happen if a model is trained on data that includes gaps or inconsistencies?
The model may learn to predict anomalies as normal, rather than recognizing them as exceptions, leading to incorrect conclusions.