• Written by Azhar Aliyar, Director

A plant manager at a chemicals facility checks the production schedule for the day, noticing a discrepancy between the planned yield and the actual output from the previous shift. This discrepancy is not unusual, given the complexities of continuous processes in chemicals manufacturing, where yield drift, feedstock variability, and plant-to-plant recipe differences can all impact production.

Challenges in chemicals manufacturing

The chemicals industry presents unique challenges for AI systems integration, particularly when it comes to connecting to systems of record without becoming a second source of truth. Yield drift, for example, refers to the gradual change in production yields over time, which can be caused by a variety of factors, including equipment wear and tear, changes in feedstock quality, and variations in operating conditions. Feedstock variability is another challenge, as the quality and composition of raw materials can vary significantly from batch to batch, affecting the final product's quality and yield.

Additionally, plant-to-plant recipe differences can also impact production, as different facilities may have slightly different processes and equipment, resulting in variations in the final product. These challenges require careful consideration when integrating AI systems, to ensure that the system can accurately account for these variations and provide reliable insights.

Read paths, write paths, and the need for care

When integrating AI systems with existing systems of record, it's essential to consider both read paths and write paths. Read paths refer to the flow of data from the system of record to the AI system, while write paths refer to the flow of data from the AI system back to the system of record. While read paths are relatively straightforward, write paths require more care, as they can potentially alter the system of record and create inconsistencies.

A shift supervisor, for example, may use the AI system to analyze production data and identify areas for improvement, but if the AI system is not properly configured, it may inadvertently overwrite existing data or create duplicate records, leading to inconsistencies and errors. Therefore, it's crucial to carefully design and test write paths to ensure that they are accurate and reliable.

Reconciling definitions that differ between systems

Different systems of record may have different definitions for the same data elements, which can create challenges for AI systems integration. For example, one system may define "production yield" as the total amount of product produced, while another system defines it as the amount of product that meets quality standards. These differences in definition can lead to inconsistencies and errors if not properly reconciled.

To reconcile these differences, it's essential to carefully map data elements between systems and ensure that the AI system can accurately translate and transform data as needed. This may involve working with subject matter experts to understand the nuances of each system and develop a common language and set of definitions. By doing so, organizations can ensure that their AI systems provide accurate and reliable insights, and that data is consistent across all systems.

Latency budgets and precomputation

Latency budgets refer to the maximum allowable delay between the time data is generated and the time it is available for analysis. In the chemicals industry, latency budgets are often tight, as production processes are continuous and decisions need to be made quickly to optimize yield and quality. To meet these latency budgets, AI systems may need to precompute certain data elements or use caching mechanisms to reduce the time it takes to retrieve and analyze data.

For example, a category manager may use the AI system to analyze production data and identify trends and patterns, but if the system takes too long to retrieve and analyze the data, the insights may not be available in time to inform decisions. By precomputing certain data elements or using caching mechanisms, organizations can reduce latency and ensure that their AI systems provide timely and relevant insights. For more information on how AsscherAi can help with latency budgets, visit our website.

Limitations and wrong choices

While AI systems integration can provide significant benefits for chemicals manufacturers, there are certain situations where it may not be the best choice. For example, if an organization's systems of record are highly customized or proprietary, it may be difficult or impossible to integrate an AI system without significant modifications or workarounds. In such cases, it may be more effective to use alternative approaches, such as manual data analysis or traditional business intelligence tools.

Additionally, AI systems integration may not be suitable for organizations that lack the necessary infrastructure or expertise to support it. If an organization is struggling to maintain its existing systems or lacks the resources to devote to AI systems integration, it may be better to focus on more pressing priorities. For organizations that are unsure about whether AI systems integration is right for them, it's a good idea to consult with experts, such as those at Harns Technologies, to determine the best course of action.