Artificial Intelligence In Manufacturing – Improving The Bottom Line
Artificial Intelligence and it’s Practical Application in the Manufacturing Environment
As the manufacturing industry becomes increasingly competitive, manufacturers need to implement sophisticated technology to improve productivity. Artificial intelligence, or AI, can be applied to a variety of systems in manufacturing. It can recognize patterns, plus perform time consuming and mentally challenging or humanly impossible tasks. In manufacturing, it is often applied in the area of constraint based production scheduling and closed loop processing.
AI software uses genetic algorithms to programatically arrange production schedules for the best possible outcome based on a number of constraints, which are pre-defined by the user. These rule-based programs cycle through thousands of possibilities, until the most optimal schedule is arrived at which best meets all criteria.
Another emerging application for AI in a manufacturing environment is process control, or closed loop processing. In this setting, the software uses algorithms which analyze which past production runs came closest to meeting a manufacturer’s goals for the current pending production run. The software then calculates the best process settings for the current job, and either automatically adjusts production settings or presents a machine setting recipe to staff which they can use to create the best possible run.
This allows for the execution of progressively more efficient runs by leveraging information collected from past production runs. These recent advances in constraint modeling, scheduling logic, and usability have allowed manufacturers to reap cost savings, reduce inventory and increase bottom line profits.
AI – A brief history
The concept of artificial intelligence has been around since the 1970s. Originally, the primary goal was for computers to make decisions without any input from humans. But it never caught on, partly because system administrators couldn’t figure out how to make use of all the data. Even if some could comprehend the value in the data, it was very hard to use, even for engineers.
On top of that, the challenge of extracting data from the rudimentary databases of three decades ago was significant. Early AI implementations would spit out reams of data, most of which wasn’t sharable or adaptive to different business needs.
The resurgence
AI is having resurgence, courtesy of a ten-year approach called neural networks. Neural networks are modeled on the logical associations made by the human brain. In computer-speak, they’re based on mathematical models that accumulate data based on parameters set by administrators.
Once the network is trained to recognize these parameters, it can make an evaluation, reach a conclusion and take action. A neural network can recognize relationships and spot trends in huge amounts of data that wouldn’t be apparent to humans. This technology is now being used in expert systems for manufacturing technology.
Practical application in the real world
Some automotive companies are using these expert systems for work process management such as work order routing and production sequencing. Nissan and Toyota, for example, are modeling material flow throughout the production floor that a manufacturing execution system applies rules to in sequencing and coordinating manufacturing operations. Many automotive plants use rules-based technologies to optimize the flow of parts through a paint cell based on colors and sequencing, thus minimizing spray-paint changeovers. These rules-based systems are able to generate realistic production schedules which account for the vagaries in manufacturing, customer orders, raw materials, logistics and business strategies.
Vendors typically don’t like to refer to their AI based scheduling applications as AI due to the fact that the phrase has some stigma associated with it. Buyers are perhaps reluctant to spend money on something as ethereal sounding as AI but are more comfortable with the term “constraint based scheduling”.
Constraint-based scheduling needs accurate data
A good constraint-based scheduling system requires correct routings that reflect steps in the right order, and good data on whether steps can be parallel or whether they need to be sequential. The amount of thorough planning that is required for a successful system to be launched is one of the largest drawbacks.
If a management team has not defined and locked in accurate routings in terms of operation sequence and operation overlap, and if it has not correctly identified resource constraints with accurate run and set-up times with a correct set-up matrix, what it winds up with is just a very bad finite schedule that the shop cannot produce. Tools like AI should not be thought of as a black box solution, but rather as a tool that needs accurate inputs in order to produce a feasible schedule that can be understood by the users.
Constraint-based scheduling within an ERP (enterprise resource planning) system
In selecting a solution, there are a number of system prerequisites that you need to look for. The better an enterprise application integrates various business disciplines, the more powerful it will be in terms of delivering constraint based scheduling. This means that if an application suite offers functionality cobbled together from different products the manufacturer has purchased, it may be harder to use that suite to deliver good scheduling functionality. This is because a number of business variables that reside in non-manufacturing functionality can affect capacity.
When an ERP package has been configured for constraint based or finite scheduling, it is generally routed to a scheduling server which calculates start and finish times for the operations with consideration to existing orders and capacity. When the shop order is executed, the scheduling system updates the information regarding operations and sends the results back to the enterprise server.
Scheduling functionality within an ERP solution ought to work in a multiple-site environment. Let’s say you need to calculate a delivery date based on a multi-site, multilevel analysis of material as well as capacity throughout your whole supply chain. The system should allow you to plan given all the sites in your supply chain and the actual work scheduled for each of those work centers. Manually or automatically, you should be able to schedule work and immediately give your customer a realistic idea of when the order will be completed.
More benefits of AI, constraint based applications
Apart from the immediately apparent capacity management benefits of constraint based scheduling, there are a number of less obvious analytical capabilities. Scheduling functionality typically allows you to conduct predictive analyses of what would happen if certain changes are made to an optimized schedule. So if a plant manager is pressured by a particular account executive to prioritize an order on behalf of a customer, that plant manager can produce excellent data on how many other orders would be late as a result. Furthermore, this functionality can provide predictive analyses on the effect of added capacity in the plant. This enables manufacturers to see if equipment purchases will truly deliver an increase in capacity, or if it will simply result in a bottleneck further downstream in the manufacturing process.