Preventing traffic jams a flexible manufacturing system
Traffic jams frequently occur when parts are manufactured on a variety of equivalent machines and have the flexibility to choose a path through the system. Just like with cars on the road, traffic jams occur for two main reasons. First, there can just be too many cars on the road; likewise part demand can exceed the capacity of the machines, tools, or transport systems. The second cause is more frustrating: a minor incident such as a car braking heavily or even a lane change in a previously smooth traffic flow can cause a ripple effect of slowdown. Similarly in manufacturing, a tool change taking a little longer than normal or a lengthy load operation at a machine can lead to a cascading failure with impact far beyond the initial hiccup. Thankfully, in manufacturing we have more control over parts and machines than in traffic engineering and can prevent traffic jams with some surprisingly simple techniques.
Capacity limitations in a manufacturing system come in a variety of forms. Just like in traffic engineering where a single intersection or on/off ramp can throttle a road with a large number of lanes, a single tool or machine or fixture can bottleneck an otherwise large manufacturing system. These kinds of traffic jams can be detected by data analysis and predicted through simulation. Once identified, these problems can be fixed by flexibility changes or hardware purchases.
Traffic jams which cascade from minor incidents are different. The total cars on the road or parts being manufactured can be under the capacity of the road or system, but a small ruffle leads to sustained traffic congestion when normal flow might have continued. In traffic engineering, these incidents are still not fully understood. Similarly in flexible manufacturing, we can detect these kinds of traffic jams using data analysis but can't identify a single cause. Instead, our strategy is to design and operate the manufacturing system so that it is robust against these minor fluctuations. By actively controlling the parts and flexibility with some simple rules, we can completely prevent these kinds of traffic jams; small hiccups will not cascade to a full blown snarl-up.
Preventing traffic jams through active management
Ramp meters, which are stop lights placed on freeway entry ramps, are a simple and effective technique to mitigate the effects of traffic congestion and decrease travel time. By limiting the rate at which cars join the main traffic flow and preventing a thundering herd of vehicles, ramp meters allow the flow to survive small hiccups and variations. In flexible manufacturing, we have had tremendous success in improving efficiency by limiting the rate at which orders enter the system. By placing a "stop light" on part demand and not entering the entire order book all at once, machines can continuously operate at high utilization.
Ramp meters can lead to frustration when not understood: "Why am I waiting on the ramp when the mainline road is free flowing?" is a common question. Likewise in manufacturing we face questions such as: "I have an order due in 4 days and you are not going to put that order into the system?" or "The machines have this 15 minute period where they are inactive, why don't we add in some part demand?" A $650,000 study (PDF) by the Minnesota Department of Transportation showed that without ramp meters, traffic throughput decreased by 14%, travel times increased by 22%, freeway travel time increased by as much as 50%, and crashes increased by 26%. These are large changes for such a simple tool! We have seen similar large improvements in flexible manufacturing; when "part order meters" are introduced, the machine utilization is typically increased by 25% to 30%.
Part order meters are incredibly effective but require buy-in from everyone. Yes, this order is due in 4 days and it is not in the system, but when we do enter that order tomorrow it will run quickly without jams or delays. Indeed, second shift produced less parts than third shift because the system was partially idle, but this is not the fault of the second shift operators and is necessary to maintain the overall throughput of the system. The most effective strategy is collective ownership of efficiency: all operators and all shift managers are jointly responsible for the overall performance of the system, and everyone understands the prevention of traffic jams is a priority.
Tactic for active management through order metering
The goal is to rate-limit the orders entering the manufacturing system. Since part manufacturing cycles are much slower than vehicles on the road, we have found that entering a chunk of orders once a day or sometimes even once a week or once a month is enough to prevent traffic jams. (The slow pace of manufacturing can sometimes be a determent since it means that when traffic jams do form, they take much longer to clear up.) The tactic is to look at the outstanding orders, decide which orders can be completed in a single day (taking into account due dates), and enter only those orders into the manufacturing system.
There are two challenges. First, it is time consuming to manually enter orders one-by-one into the system and even the easier technique of holding and unholding parts still dumps a large chunk of demand into the system and leads to traffic jams. Most control computers and machine software have some form of order import via files, spreadsheets, or via code, but each machine software is different and periodically preparing the import file in the right format is still a challenge. Our SeedTactic: Order Link software works with a range of automation machine company's software out of the box. Also, since Order Link automatically connects to your ERP system, the transfer of orders from ERP to the machine can happen without a time-consuming manual process.
The second challenge is accurately predicting throughput. Depending on the mix of parts and cycle times it can be difficult to project which orders can be completed in a single day. In traffic engineering, each car or even semi trailer takes up about the same amount of road capacity so the timing of the red light on the ramp meter can be set independently of the specific vehicle waiting to enter the freeway. In manufacturing, we have a large variety of order sizes and part capacity requirements which change day-by-day, so how do we decide which orders should receive the green light? An additional complication is that typically only a portion of the orders are flexible (for quality control and ease of part programming), so enough orders must be entered to ensure the system is kept busy.
Our SeedTactic: Order Link uses the flexibility plan to accurately and continuously compute which orders can be completed and which orders are necessary to keep the system full but not overloaded. This allows OrderLink to automatically respond and adapt to any mix of orders or due dates, without any pre-planning or customized logic. The OrderLink software even applies some artificial intelligence to adapt the flexibility (e.g. change which parts can use which lanes) to maximize utilization and simplify inspection and quality control.
TL;DR
Traffic jams in flexible manufacturing systems decrease machine utilization by 20%-30% and occur with frustrating regularity, especially in systems without detailed monitoring and tactical prevention of order congestion. These traffic jams can be prevented by limiting the rate at which orders enter the system, similar to stop lights on freeway entrance ramps. Our SeedTactic: OrderLink software does exactly that: it accesses your outstanding orders, computes which orders can be completed in a single day (while taking into account order due dates), and copies those orders automatically into the manufacturing system.