Artificial intelligence for real-time manufacturing execution & operations management

Feb. 1, 2017

AI makes systems less costly and easier to use for small-batch, quick-turn process manufacturers

Many mid-sized process manufacturers, such as those making paints, cosmetics, coatings, food and beverages, and over-the-counter pharmaceuticals, are facing increasing demands from their customers for small, customized batches of product delivered rapidly, often with only a few days to turn around the order. This is in contrast to their previous operations, which created large volumes of standard products with production runs measured in weeks or months.

This change is being driven by increasing consumer expectation for products that are tailored to their specific needs, often ordered over the internet and delivered “overnight” by package delivery companies in combination with the demand for “lean” inventories throughout the supply chain.

The result is a dramatic increase in the complexity of operations management – tracking, planning and scheduling operations when orders are essentially arriving in a random stream with fulfillment times in days if not hours. It is no longer possible to plan and schedule production using standard operations planning methods, based on monthly sales forecasts and production plans. Instead, managers are forced to use real-time planning and scheduling methods, especially when equipment goes down, critical employees get sick or the truck with an essential delivery of materials ends up in a ditch.

To handle this complexity, manufacturers are increasingly turning to the use of real-time manufacturing execution systems (MES) in combination with real-time manufacturing operations management (MOM) systems to track and manage their operations. They are finding their legacy enterprise resource planning (ERP), materials requirements planning (MRP), and MES systems, used in combination with planning boards and spreadsheets, are no longer able to meet the tracking, planning and scheduling needed in today’s real-time, demand- driven environment.

This article looks at how the use of real-time artificial intelligence (AI) methods can dramatically simplify and lower the cost of deploying these real-time MES and MOM systems, making them affordable and usable by small and mid-sized process manufacturing organizations.

Issues with transitioning to a real-time manufacturing system

Automating the tracking, planning and scheduling of manufacturing operations using a MES combined with a real-time MOM system has many operational benefits including:

  • Giving real-time manufacturing visibility of the status of production, inventory and customer orders
  • Preventing operational mistakes
  • Eliminating unnecessary overhead labor costs
  • Helping to prevent late shipment of customer orders
  • Enabling “lean” management of inventory
  • Automatically collecting materials’ traceability data
  • Increasing sales by offering reliable delivery of semi-custom products within short delivery times

Many manufacturing organizations, however, investigate using MES systems with technologies such as barcode tracking but then choose to stick with their old methods of using paper forms, manual keyboard data entry and planning boards, along with a small army of expediters and other employees who struggle to get orders out on time.

Some of the reasons given are:

  • The technology is too difficult or complicated to adopt.
  • Using barcode scanning will slow down workers.
  • Integrating an MES system with other systems is too hard.
  • It is too disruptive to operations.
  • It is too expensive.

These reasons are given even though market pressures for ever-shorter delivery times of semi-custom products in ever-increasing volumes at ever-lower costs make using paper forms and manual tracking methods more and more difficult.

As one expert in the field stated, “My clients are too busy drowning to learn how to swim.”

AI to the rescue

Embedding AI into an MES/MOM system means integrating rules (often user defined) and algorithms (with user-defined parameters) into a system to make it much smarter. In doing this, manufacturers can harness the power of the computer to rapidly analyze large volumes of data in real time and make routine decisions without the need for human intervention.

Contrast a Ford Model T with today’s next-generation, AI-based cars. The Model T had manual starting, manual steering, manual braking, manual transmission and no safety features. Today, we have “self-driving” cars with 26 on-board computers and embedded AI to provide push-button starting, automated braking, automated parking, the ability to drive itself down the highway, and many accident-prevention and safety features.

In a similar way, embedding AI into real-time MES/MOM systems makes them much easier to use.

Some of the ways that AI makes these systems easier to use include:

  1. Minimizing data capture time by only capturing the minimum data needed. Instead of presenting a user with 20 or so boxes on each screen and expecting the user to know which boxes to fill out (typical of ERP and accounting systems), the AI system presents the user with a form and only a minimum number of data entry boxes. The system then adjusts the form as data is entered based on rules embedded in the system.
  2. Preventing mistakes. As each data item is entered, the data is checked by rules and algorithms, and the user is warned immediately if he or she is about to make a data entry or operational mistake to allow for corrective action, just like a self-driving car does. This minimizes training time.
  3. Enabling managers, supervisors and customer support personnel to keep track of what is going on in real time. The data captured are interpreted by the built-in rules and algorithms as a real-time manufacturing status of inventory, jobs, customer orders, people and machines that can be seen anywhere and anytime there is an internet connection. This data can be shared with internal employees as well as selectively shared online with customers and vendors. This dramatically reduces the amount of time everyone spends trying to make sure customer orders get shipped on time and to answer those “What is the status of my order?” telephone inquiries.
  4. Intelligently exchanging information with systems used by other departments. There is no need for operations managers to have access to all the data used by accounting, human resources or sales. But selective data exchange, controlled by a set of rules specific to each organization, can enable managers, supervisors and customer support employees to do their jobs much more efficiently.
  5. Automatically scheduling jobs through a production facility based on rules and parameters set by the production manager. This eliminates the need for managers to continually answer those “What do I work on next?” questions. Instead, the AI-based algorithms make the myriad of routine real-time manufacturing scheduling decisions that are needed in a modern manufacturing plant. However, managers can always adjust the schedule when needed to account for special circumstances.
  6. Assisting managers in deciding what jobs to schedule and what materials to order. An AI-based system is able to predict when inventory shortages will occur in response to a continual inflow of new customer orders. The AI-based system can then alert managers when these problems are about to occur and assist them in analyzing the data to make smart decisions.
  7. Alerting managers when problems are about to occur. The AI-based rules and algorithms continuously scan the data collected by the operations tracking system, as well as data imported from other systems, to find situations that may be problematic. This may be as simple as a job step taking too long to complete or a prediction that a customer order will not be shipped on time. The system is then able to send alerts to managers in the form of email messages or text messages sent to their mobile phones so they can take immediate action. This saves them from “walking the floor,” being “glued to a computer monitor” or poring over reports to try to determine when something is about to go wrong.

AI makes MES/MOM system deployment easy & inexpensive

A common response when companies are introduced to the use of real-time manufacturing with AI-based MES/MOM systems is: “I can see the advantages, but we are just a small organization. Real-time AI sounds awfully complicated, and we only have one [or no] IT guy.”

The reality is, if you can fill out a spreadsheet, then you can quickly set up the rules to tailor a system to meet your specific needs.

Using remote installation and AI-based, real-time MES/MOM software available today, most companies can typically have a real-time manufacturing operations tracking and management system up and running within a couple of hours. After that, filling out the spreadsheets used to set up the knowledge base typically takes a few hours to a few days, depending on the complexity of the manufacturing and warehousing process.

Remote training time on how to use the software typically runs from two to six hours, again depending on the complexity of the manufacturing business. These times extend somewhat when clients need to set up the rules about how to exchange data with other systems or to print barcode labels in custom formats. But even here, the time needed is measured in hours or days, not weeks, since the software provides or automatically generates most, if not all, of the instructions needed for these intelligent data-exchange interfaces.

An AI-based, real-time MES/MOM system is also relatively inexpensive. Monthly fees for renting the software for installation on a client’s own computer are around $1,000 per month for a small- to mid-sized manufacturing operation. This is a fraction of the $4,000 to $8,000 per month cost of each person, who is now no longer needed to perform overhead tasks such as manual data entry, expediting, planning and scheduling of manufacturing operations.

Contrast this to setting up ERP systems with embedded MES capabilities in which deployment times of months, if not years, are typical, and in which costs are into hundreds of thousands of dollars.

AI in the future

AI-based, real-time MES/MOM systems have been available for over a decade and initially required a fair amount of expertise to configure. Today, it is increasingly common to see companies use these software tools and methods to do most, if not all, of the deployment of these systems themselves.

As a result, companies ranging from small food processors to large plastic recyclers have been able to rapidly scale their businesses by taking advantage of the demand for semi-custom products delivered rapidly in small batches when needed by customers.

Future endeavors for AI-based, real-time MES/MOM software solutions include current projects that do “deep learning,” using the thousands of processors available on today’s low-cost graphics processing cards. By automatically learning about materials processing, delivery and wait times, based on many operational parameters, the system may be able to answer questions like, “When will my order be delivered?” This information would then be used to do automated real-time manufacturing scheduling and planning and alert managers when the system predicts an order may be shipped late.

This deep learning will make deploying these real-time MES/MOM systems even easier by eliminating the need to have operations managers set up scheduling and planning rules. Instead, the system will use model-based reasoning to learn the parameters of an operations model. It will do this based on neural network processing of the large volume of real-time manufacturing data generated by barcode and mobile data collection devices as well as increasingly through automated radio-frequency identification and Industrial Internet of Things sensors.

Dr. Peter Green received his Bachelor of Science in electrical engineering and a doctorate in computer science from Leeds University in England. He was a senior member of the research staff at MIT where he performed research into real-time intelligent systems under a Defense Advanced Research Projects Agency (DARPA)-funded contract. He was subsequently a full professor of computer engineering at Worcester Polytechnic Institute where he performed research into software methods for implementing real-time, intelligent-agent-based systems for the U.S. Air Force and NASA. He then founded BellHawk Systems to continue this development using Small Business Innovation Research grants to enable commercialization of this technology. During the past decade, Green and his team applied these real-time AI methods to the development of real-time manufacturing operations tracking, scheduling, planning and mistake-prevention systems for manufacturers. Green may be reached at [email protected].

BellHawk Systems

Sponsored Recommendations

2024 Manufacturing Trends — Unpacking AI, Workforce, and Cybersecurity

The world of manufacturing is changing, and Generative AI is one of the many change agents. The 2024 State of Smart Manufacturing Report takes a deep dive into how Generative ...

The Journey to Operational Excellence: Quality-Driven Compliance

Learn firsthand from top industry analysts how to navigate regulatory compliance (i.e. FSMA) & food safety audits in manufacturing.

Cold Chain Tracking with FactoryTalk PharmaSuite

Manage thermo-sensitive materials, optimize production & reduce waste

State of Smart Manufacturing Report Series

The world of manufacturing is changing, and Generative AI is one of the many change agents. The 2024 State of Smart Manufacturing Report takes a deep dive into how Generative ...