Andreas Eschbach, CEO and Founder
Why is that emulsion suddenly too thick? What’s the optimal maintenance interval for the reactor? Has the centrifuge ever acted in this way before, and what was done to fix it? There’s a lot of hidden information in historical shift logs, maintenance records, and plant data—if you know how to find it. An AI-enabled Smart Search system trained on your specific terminology, workflows and user requirements could help.
Process manufacturing plants in the chemical or pharmaceutical industries generate terabytes of data, including both automated sensor and equipment readings and human-generated logs and observations. Plants are increasingly centralizing all this data in digital Plant Process Management (PPM) systems to enable efficient data analysis and knowledge transfer.
The sheer volume of
data contained in these systems, along with the highly technical nature of the
data, makes information retrieval a challenge. Most PPM search systems still
rely on simple keyword search. That means that users must know the exact search
term needed to find the information they need. It also depends on entries being
complete, correct and properly tagged and formatted. If I don’t know exactly
what I’m looking for, or if the entry I really need has incomplete information,
the system is not able to help me. Alternatively, it may return thousands of
potential results for a common keyword, leaving the user to sort through the
list to find the one closest to what they are looking for.
That’s where artificial intelligence (AI) can help. While traditional keyword search depends on exact matches to keywords in its database, AI-enabled Smart Search uses advanced algorithms to generate more relevant and context-specific results.
Smart Search leverages AI technologies such as Natural Language Processing (NLP), machine learning (ML) and semantic search to understand the user’s intent, the context of the query, and the relationships between words. This results in a more sophisticated, accurate, and relevant search experience. Instead of just looking for keywords, the algorithm understands the meaning of words and can expand search to synonyms and related keywords based on that understanding.
Smart search can understand and respond effectively to complex or ambiguous queries and queries framed in natural, conversational language. For example, instead of trying to guess the right keyword, they can simply ask, “Why is this product brown instead of grey?” The Smart Search system can parse the question to understand the user’s intent and context, search through terabytes of structured and unstructured data to find relevant results, and rank results according to their predicted usefulness. Moreover, AI-enabled smart search can learn from past search patterns and behaviors to deliver increasingly relevant results over time.
This makes Smart Search a powerful tool for dealing with complex queries or large, diverse sets of data. It turns a PPM system into an engine for knowledge management and transfer that can drive significant operational improvements.
Many of us are already using AI-enabled Smart Search tools available to the public, such as Microsoft’s Bing or Google’s Bard. These tools, while far from perfect, are transforming the way many of us search the web. Instead of visiting individual websites returned by keyword search to see which one has what we’re looking for, we can use these emerging AI tools to ask complex questions and get answers synthesized from multiple sources across the web.
While useful and entertaining, these general-purpose Large Language Models (LLMs) still have a long way to go in terms of information accuracy and suitability in a search context. Smaller, more specialized AI search tools have been created to assist search in specific domains, such as library systems, ecommerce sites, media platforms or helpdesk applications. There are now a number of off-the-shelf AI solutions available that can be integrated into various platforms and systems.
Process manufacturing, however, is a highly specialized application that is not well suited for off-the-shelf solutions. The chemical and pharmaceutical industries—and individual companies within the industries—have unique terminology, abbreviations, workflows, and systems that a general-purpose AI will not understand. To get the maximum benefits of Smart Search, the system needs to be trained in industry- and company-specific data and language.
That’s what eschbach did for one of our clients in the agrochemical industry. We conducted a systematic customer research study with user groups, workshops and onsite investigations to better understand their workflows, language and user requirements. Working in close collaboration with the client and AI experts from Göttingen University, we created a company-specific Smart Search program tailored to their processes, systems and user requirements.
A Smart Search system trained on domain-specific data and tailored to the needs of a particular organization is a powerful tool for information access, knowledge transfer and operational improvement. Using the Smart Search system, employees of the agrochemical company can quickly find the information they need in the 8+ years of historical data they have in eschbach Shiftconnector.
Critically, the Smart System is able to understand the context of complex queries and find relevant results even if entries are incomplete. For example, users can find potential solutions to problems in historical data, even if problems and solutions were incompletely described or missing keywords. Providing more efficient access to historical information accelerates troubleshooting and problem-solving and improves decision-making across the organization. Smart Search has reduced the time it takes users to find answers to complex queries from several minutes (or longer) to just seconds.
Smart Search is the next stage of digitalization for Shiftconnector users. When people can find the information they need, at the time that they need it, they are empowered to make better operational decisions. That’s how we turn historical data into actionable knowledge that drives performance improvement across the organization.