When an agrochemical company needed a better way to help workers find historical knowledge, they turned to artificial intelligence (AI). An AI-enabled Smart Search system helps employees find usable information and trends in digital data from shift logs, maintenance notes, equipment readings and more. Better access to historical data speeds up troubleshooting, enables better decision making and empowers employees at all levels of the organization.
The Problem: Knowledge management in processing plants
Process manufacturing plants generate thousands of data points every day, from automated sensor readings to notes and observations from line workers, maintenance staff or engineers. Many processing plants have invested significant resources in digitizing systems that have traditionally been offline, such as shift handover, inspection rounds, maintenance logs, and ad hoc communication between individuals and teams. At the same time, Industry 4.0 applications and Industrial Internet of Things (IIoT) devices now generate massive amounts of automated data, including sensor and equipment readings.
All this data has the potential to be tremendously useful for problem-solving, troubleshooting and decision-making. The answers to today’s problems are often hidden in mounds of historical data — if you can find them. For example, if an employee notices a change in product color or consistency, it can be extremely helpful to know if the issue has ever been observed before and how it was resolved.
However, finding the right information, at the right time, can be a significant challenge. Process manufacturers face many problems that make it difficult for employees to retrieve relevant information when needed.
- Information may be spread across multiple systems and channels, including the manufacturing execution system (MES), enterprise resource planning (ERP) systems, manually updated shift logs and inspection notes, and emails or Slack channel communication between employees.
- Manual entries are often incomplete or contain misspellings or plant-specific terms and abbreviations.
- Some knowledge is offline or even completely tacit — that is, largely held in the heads of experienced employees.
- Employees may not know exactly what information they need to search for or where they need to look to find it.
Digital tools for shift handover, inspection rounds and Plant Process Management (PPM) can help manufacturers collect and centralize data and notes for future retrieval and analysis. A centralized knowledge management platform acts as a historical archive, collecting both formal data and tacit knowledge that is often hiding in shift logs or maintenance notes. However, just collecting information is not enough. Employees also need an efficient way to retrieve it.
Most PPM systems rely on basic keyword search for information retrieval — but keyword searches don’t always generate the best results.
- Keyword search depends on entries being complete and properly labeled, with all words spelled out correctly. A keyword system is not able to extrapolate from incomplete information.
- Users must use the precise keywords to return a relevant result. If they don’t know the right term to surface the information they are looking for, it remains hidden.
- Common keywords may return thousands of results, which will not necessarily be offered in the order of highest relevance. Users may have to sort through many pages of search results, or use increasingly esoteric combinations of terms, to find relevant information.
These were some of the problems faced by the agrochemical company. Digitizing shift handover and PPM activities created a rich historical archive with many years of collected data. However, it was difficult for employees to find relevant and usable information when it was needed. That’s where AI can help.
Search vs. Smart Search: What’s the difference?
To solve the knowledge management problem, the agrochemical company implemented a Smart Search system for their PPM. Smart Search utilizes AI tools to enable more efficient search and return more relevant results for the user. A Smart Search system uses two powerful AI tools to assist users.
- Natural Language Processing (NLP) is a form of AI that allows it to understand human language. It allows the system to understand user queries given in plain language. For example, a user searches for a viscosity problem. Not all users will search the same way. Some may search that “It flows like honey” while another may search that the “raw material was like cream.” All these semantic similarities can be found with one search prompt using NLP from a Smart Search. The algorithm can understand what the user is looking for and find the results that are likely to be most helpful.
- Machine Learning (ML) is a form of AI that allows the algorithm to find patterns in and learn from large datasets. For instance, over time, as users search for and select specific results, the ML algorithm recognizes these patterns and preferences. This self-improvement means the search becomes increasingly refined and efficient, adapting to user behaviors and preferences.
A Smart Search system can vastly improve search efficiency and connect people to the information they really need. In turn, this enables more efficient and effective problem-solving and decision-making.
In a highly specialized domain such as agrochemical manufacturing, an off-the-shelf AI solution is not adequate. The AI has to be trained on domain-specific language, including technical terms and abbreviations specific to the industry or even the company. To improve the efficiency of the Smart Search system, you need to adapt an AI search tool for their needs.
Power to the people (with AI)
A Smart Search system empowers employees by connecting them to the knowledge of the past and giving them the information they need to perform their jobs effectively.
At the agrochemical company, the Smart Search program has substantially reduced the time it takes to find relevant information in a search from several minutes to just seconds. Now, they can easily search through more than eight years of digitized logs and data and find the answers they need. With NLP, results don’t depend on workers knowing the exact search term or entry logs being correct and complete. The system has been well accepted by employees at all levels in the company, from shop-floor operators to process engineers.
AI-assisted search can help process manufacturers move beyond simply digitizing data and into the next stage of digitalization. It’s part of the move from Industry 4.0 (digitization and automation) to Industry 5.0: using technology to empower people.
When people have instant access to the information they need, they can make better decisions, stay more productive, and quickly resolve emerging issues. As novel issues come up and are resolved, that new knowledge is also added to the database for future reference. A centralized knowledge platform with Smart Search can be an important element of knowledge transfer for less experienced employees, ensuring that the collective wisdom and knowledge of the experienced workforce are not lost. Better knowledge management will, in turn, make tomorrow’s processing manufacturers safer, stronger and more resilient.
Andreas Eschbach is the founder and CEO of the software company eschbach, which helps production teams stay safe and work smarter through better information sharing and collaboration. Holding a degree in computer science, he draws his practical experience from leading a variety of international software consulting and implementation projects for leading chemical manufacturing companies, focusing on production, continuous improvement, EHS and maintenance. His company is a provider of manufacturing solutions and headquartered in southern Germany and has an office in Boston, Massachusetts.
eschbach
