Skip to content

Flexible job shop scheduling

Three machines can run the operation, so you choose machine and start time. Our model chooses both together and spreads work across all three.

  • Free 30-minute call
  • No commitment
  • You own the code
Typical industries
Manufacturing
Problem type
Scheduling
You receive
A documented model, run on your data

In short: A flexible job shop schedule plans a workshop where an operation can run on more than one machine. It decides both at once: which machine runs each operation, and in what order. That way jobs finish sooner and the work is spread evenly across the machines.

The problem

Monday, seven in the morning, and you’re working on the machine schedule. Three milling machines can all run the next operation. You send it to the fastest one, as always. By Wednesday that machine has a queue, the other two run only half the time, and two orders are late.

Signs the shop loses time between machines

The fastest machine always has a queue, and the slower ones wait for work. You assign jobs to machines by rule of thumb. Then you set the order of jobs on each machine separately. When a machine goes down, the plan no longer works and you redo it by hand. You miss delivery dates, even though the shop seems to have enough free machine hours.

Why assign-then-sequence fails

In a classic job shop, each order takes its own route through the machines. Each operation, such as drilling or milling, has one fixed machine. In a flexible shop, an operation can run on any of several suitable machines. So for every operation that is next in line, you face two questions at once: which machine, and when? The usual approach answers them one after the other.

The usual approach first assigns every operation to a machine by rule of thumb, for example the fastest machine or the one with the least work. Then it sets the order of jobs for each machine separately. Each step feels manageable. But when it picked the machines, it didn’t know what the order of jobs would look like. A machine that looked free is overloaded by the time the work arrives. The machine choice is fixed, so a different order can’t fix it. Work piles up at one machine while others wait, and orders leave late.

What changes

Today, each operation goes to the fastest machine, and the queue builds there while the other two machines wait for work. With our model, you get machine and timing for every operation, decided together. Some jobs go to a slower machine that’s free, instead of queuing for the fast one. You check the plan and move a job if a machine goes down.

Machine assignment scheduling optimisation: what you get

For every operation, the model picks the machine and the start time so that all orders are finished as early as possible. Work spreads across all machines that can do it. So no single machine becomes the bottleneck that holds everything up. When a rush order comes in or a machine goes down, the model recomputes the plan on the same logic. You no longer give up a weekend for the machine schedule.

We build the model on your own routings, the steps each order goes through, plus your processing times and machine availability. We test it on your own shop-floor data before it goes live. We pick weeks with rush orders and machine breakdowns, because that’s when assign-then-sequence loses the most time.

  • A model written around your own flexible job shop scheduling rules.
  • Results measured in the numbers you already track.
  • It slots into how you already work, and you can check every answer it gives.

How it works

Your jobs and their operations go into one model. An operation is one step of a job, such as drilling or milling. For each operation, the model knows which machines can run it and how long it takes on each. It decides machine and start time for every operation together. So it doesn’t send each operation to the fastest machine, where the work then piles up. You get the plan that finishes all jobs soonest and keeps every rule about machines and order. You check it before it goes to the shop floor.

What the FJSP is

The formal name is the flexible job shop scheduling problem (FJSP). In a standard job shop, every operation is tied to one machine. The only decision is sequencing: the order of operations on each machine. When several machines can run the same operation, a second decision appears. That’s machine selection: which machine runs each operation.

Each operation has several machines that can run it. Each machine has its own speed, queue and free times. FJSP combines two problems: assigning operations to machines and putting them in order. That makes it harder than either one alone. It’s also the version that matches how most shops run in practice.

A four-job example of why “fastest machine first” loses

Take two machines that can both run the same operation. Machine A needs 2 hours per job, machine B needs 3. Four identical jobs arrive at once. The rule “fastest machine first” sends all four to A. The last one finishes after 8 hours, while B has no work the whole time.

Send three to A and one to B instead. A finishes after 6 hours, B after 3, and the last job leaves at hour 6. That’s two hours earlier, on the same machines. In your shop, routings, due dates and setup times come on top. A setup time is the time you need to prepare a machine for the next job. But the pattern stays the same: the best machine for one job is often the wrong one for the schedule.

A smarter rule, “the machine that finishes this job soonest”, solves this toy case. It stops helping once jobs have several steps on different groups of machines. Then one machine choice can delay a later step on another machine. A rule that looks only one job ahead can’t see that.

Makespan and workload balance

Two goals matter most. The first is makespan, the time from the start of the first job to the end of the last. A shorter makespan means more orders out of the door each week. If you build only to customer order, that often decides whether you hit a delivery date.

The second is workload balance across machines: every machine should have a similar amount of work. With a balanced load, no single machine becomes the bottleneck. It often gives the shortest makespan too. Workload balancing in a job shop depends on the machine choices as much as on the sequence. An assignment can look ideal on paper and still be impossible. That happens the moment the chosen machine is down for maintenance, waits for a tool or has no operator.

Joint optimisation: MILP, constraint programming or metaheuristics

Joint optimisation decides machine choice and sequence as one problem. Which method fits depends on the size of your shop. With every job and machine you add, the number of possible plans multiplies. FJSP belongs to a class of problems called NP-hard. That means no known method solves every large case quickly. For medium-sized shops, mixed-integer linear programming (MILP) gives a provably good plan. It also tells you how far, at most, the plan is from the best possible one. MILP writes the shop as equations with yes-or-no choices, such as which machine runs an operation.

Constraint programming suits this problem well. The model first writes every rule of your shop down as a condition. Then it searches for a plan that keeps them all. Google’s free optimisation software OR-Tools (OR stands for operations research) includes a worked flexible job shop example. For large shops or fast replanning, metaheuristics take over. These are search methods that improve a plan step by step. Examples are tabu search and large neighbourhood search. When the time you set runs out, you get the best plan found so far.

Where flexible job shops show up

In manufacturing, flexible shops are common. Computer-controlled machining centres (CNC, computer numerical control) can run many kinds of operations. Many shops also have groups of machines that do several kinds of work, and workstations that can replace each other. So you have to choose a machine many times a day. You get a custom FJSP algorithm in place of a dispatch rule, a fixed habit like “fastest machine first”.

Common questions.

Something else? Ask us directly.

No. More machines of the same kind is still a classic job shop. The flexible case means an operation may run on several different machines, each at its own speed. That adds a second decision, which machine, on top of the sequence. That’s why simple dispatch rules, fixed habits like “fastest machine first”, lose time there. A schedule that sets machine and sequence together wins that time back.

You set the time. The search methods for this problem return the best plan found so far whenever you stop them. You calculate the plan for the next day overnight, so the calculation may take several minutes. When a machine breaks down during the day, you need a new plan fast. Then you give the search far less time and accept a good plan over a proven best one. We tune that trade-off on your data.

The model replans from the current state of your machines and jobs. Operations already running stay where they are, and the rest are reassigned across the machines still available. The next hour or two can stay unchanged in the plan, so the people at the machines don’t get a new plan every few minutes. You see what moved and why, and confirm it.

Sometimes it can, and if that’s the case, we’ll tell you. Standard software is built for an average operation, though. Yours runs on its own rules: your contracts, your exceptions, the things you know by heart. A custom model writes those rules in, and that’s where the better plan comes from.

You don’t need either to start. We build the model from the data you’ve already got, even if it lives in three spreadsheets and one person’s head. You get a system you run yourself. We train your team until they no longer need us.

You see a first result with the prototype, before you commit to anything bigger. We start with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers.

Tell us what you’re planning.

In a free 30-minute call with a specialist, you find out whether optimisation can help.

  • Free, no commitment
  • You talk to a specialist directly
  • A clear go or no-go