Using AI in Industrial Digitalization
In an earlier blog post we discussed how we at DSS Consulting use ChatGPT in general, how it helps our development efforts. Let us now demonstrate the use of AI in industrial digitalization through a specific project.
The task at hand is to provide a useful addition to the EffiSensy solution in the ongoing IPAMS (Industrial Process Analysis and Monitoring System) project, conducted jointly with Bosch Rexroth.
EffiSensy, in short, is a solution to support efficiency on the production line, measuring and analyzing three key factors of efficient production: the number of pieces produced, the scrap rate and the availability of machinery.
Our goal is to complement the EffiSensy dashboard, which communicates with the database to answer user questions: Based on the user’s question, it writes and runs an SQL query and provides an answer–previously in tabular form, now in text.
AI in industrial digitalization–what’s the point?
The key benefit for the user is not having to browse the EffiSensy dashboard (or spreadsheets) if he or she has one or more specific questions as the software gives a specific text answer to a text question. Such a question could be, for example:
- Which machine line generated the most scrap last week?
- Which machine had the most downtime yesterday?
- Which operator produced the most product last month?
The first version of our software simply generated a SQL query for the user question via the OpenAI API using GPT 3.5 Turbo. The second version, currently under development, goes further: It now acts as an “agent”, it can “think” we have also developed additional tools to help it generate SQL queries. The agent uses a few-shot technique to arrive at the best answer in several steps, using tools, one or more of which can be omitted (as per the agent’s “decision”).
OpenAI’s recent development announced on November 6 has posed a major challenge: LangChain has been a useful tool, but the new development has made its use largely unnecessary, requiring a new logical approach.
The next step is therefore to replicate the new OpenAI development in further refining the EffiSensy add-on: this is where we stand today.
Could we help your company with similar AI-based solutions? Why not talk about it over a good cup of coffee?
