Successful plant managers must have constant visibility into several key performance indicators (KPIs) to make informed business decisions quickly and effectively. In the past, plant managers have relied on paper reports to help in their decision making. But, by their very nature, these reports are limited since they lack flexibility and can only provide static data.

Having embraced digital transformation, others have switched to dashboards in desktop computers at their offices. While these dashboards are flexible and powerful, they still require access to a tethered device.

However, many times, decisions are made at the floor. Therefore, plant managers must have the tools that allow them to get the information that they need, when they need it, wherever they happen to be.

GErryBot was designed as a product extension to GE's Enterprise Plant Insights product to quickly answer questions and provide deeper insights into the manufacturing process. Since GErryBot is available through web, desktop app or iPhone app, plant managers do not need to be at their desk or have reports in front of them to make critical decisions.

What it does

GErryBot increases the visibility into real-time production and plant operations with an easy-to-use chat interface. Rather than waiting for reports or sifting through multiple layers of a dashboard, decision makers can ask GErryBot for information about what is currently happening at the plant. GErryBot provides insights into manufacturing's key performance indicators (KPIs): yield, products completed, products scrapped, percent complete, work in process (WIP), WIP percent, opened defects, defects per unit (DPU), lead time, cycle time, and idle time. Details by operation are available for: yield, defects and WIP.

Since GErryBot is available in Slack through web, desktop app or iPhone app, the site leaders do not need to be at a desk or have reports in front of them to make critical decisions. They can quickly get answers to:

  • How many defects are open?
  • What is the cycle time?
  • What was the yield for product C9625529T on 6/2/17?
  • How many were scrapped yesterday?
  • What is the percent complete today?
  • How long will it take to produce D9625362W?
  • What is the WIP percent?

How we built it

We used some of the awesome emerging technology that is enabling the digital transformation in Industry 4.0: GErryBot Architecture

  • GE's Brilliant Manufacturing APIs - IoT data from plant manufacturing process is made available through the Enterprise Plant Insights APIs in real-time
  • Slack - Collaboration tool for teams allows extending the functionality through the creation of apps and bots. We used Slack as the user interface to host the bot.
  • Amazon Lex - Natural language understanding is done using the Artificial Intelligence engine behind Amazon's Alexa to recognize user's intent.

We built the Analytics engine to process the questions from the users, obtain the data from the Brilliant Manufacturing APIs, build visualizations, and craft a Slack message as a response.

GErryBot was not created by, affiliated with, or supported by General Electric, Amazon Web Services or Slack Technologies, Inc. and their affiliates.

Challenges we ran into

Integrating many new, emerging technologies was an interesting challenge. Particularly, finding documentation that would explain error messages was hard as the documents assume that everything will work right the first time. The best way to learn was to experiment. We also made extensive use of debugging logs.

Accomplishments that we're proud of

We brought out insights from manufacturing data by making use of features available to us:

  • Obtain and display real-time data - All the data shown is obtained from the GE's Enterprise Plant Insight's APIs in real-time. When the sample factory is "in operation," data for the current day is refreshed every time a user asks a question. Ask: "How many products have been scrapped today?" throughout the day and you may get different values each time.

  • Highlight relevant issues that may need attention - We formatted the text data to bring attention to areas of concern. In particular, when asking for the details by operation we highlighted in bold: lowest yield operation, operation with highest defects per unit (DPU), operation with highest (total) work in process (WIP). In charts, we marked in red operations with lowest yield and lowest WIP percent in production. A quick glance and users can easily find these highlighted areas in the data. Ask: 'Get defect details" to quickly learn which operation had the highest DPU.

  • Visualize plant operation data - A picture is worth a thousand words. We felt it was important to integrate charts to really brings out insights. The Slack integration allowed us to add visualizations to text messages. Ask 'What's the yield?" and then select 'Get Yield Detail' to see 2 types of visualizations used. Get Yield Get Yield Detail

  • Provide inputs through buttons and guided prompts - No one likes entering long ID strings in a mobile device! When starting the interaction, we provide buttons to select Product IDs. Also used buttons to help the user with next actions. Ask: 'What is in production?' to see a list of the product IDs. Then ask, 'How many are in process?' and you will be asked to select a product ID.

  • Ask questions in natural language - No need to learn a query language to get to the data. Users can ask relevant questions in English - typos and all! If GErrybot doesn't understand, ask for help.

What we learned

The GE Brilliant Manufacturing Suite provides a rich set of APIs to move manufacturers forward in their digital transformation. The APIs have data regarding many aspects of operation that facilitate the decision making process by having the relevant information accessible in real-time.

What's next for GErryBot

Using the data from GE's Enterprise Plant Insights, we envision expanding the capabilities of GErryBot to include:

  • Additional data from the Brilliant Manufacturing Suite of APIs
  • Smart alerts that will direct the message to the relevant user or group when a threshold has been reached
  • Interactive reports to let users zoom in to the data of interest
  • Trend analysis to set dynamic thresholds
  • Predictive analytics to anticipate issues before they happen
  • Machine learning capabilities from the Predix platform

In addition, with a conversational interface, we hope to improve the training of the bot from users. So, give it a try!

Trying it out

GErryBot has been integrated into Slack - a messaging app for teams available on the web, Mac desktop app, Chrome shortcut for Windows or as a free iPhone app - just search for “Slack app” in the iTunes store.

Get an invite to our team’s Slack channel: DataCrunch Lab Bots. You will get an email with a link to create an account. After signing in, add (+) @gerrybot to your DIRECT MESSAGES list to chat privately with GErryBot.

Built With

  • amazon-lambda
  • amazon-lex
  • ge's-enterprise-plant-insights-api
  • python
  • slack
Share this project: