How to Estimate the Real Throughput of a Bending Area with Short Batches ART-021

Bending-area throughput is not the number of bends per minute

Short-batch production is easy to evaluate incorrectly if you look only at the speed of the press brake itself. The machine may complete its working stroke very quickly, but between two batches it is not producing finished parts: the operator changes tooling, loads the program, prepares blanks, checks the first part, clears space for finished work, or waits for the next job.

For a bending area, a more useful question is: how many accepted parts pass through the entire resource over calendar time, and what consumes that time?

In a large repeating run, setup is spread across hundreds of identical parts. In a short batch, the same changeover may happen many times during a shift. That is why two areas equipped with identical presses can have very different throughput.

You do not need a universal benchmark for “correct utilization” to evaluate this. You need your own time model showing what happens between the arrival of a batch that is physically ready for bending and the exit of its last accepted part.

Count batch time from the previous job to a completed set of accepted parts

The most useful unit of analysis is not one bend or one part, but a batch as a completed production job.

For batch `i`, you can use a simple accounting model:

`T_batch_i = T_setup/changeover + T_first_part + T_repeat_bending + T_blocking_moves + T_blocking_waiting`

This is not a normative formula. It is a way to keep different kinds of time from disappearing inside one average.

Setup/changeover is the change or rearrangement of punches and dies, preparation of the required station, program change, and other work without which the next job cannot start.

First part is the actual time from the start of a new job until the process is confirmed to produce an acceptable result. If geometry, material, or tooling changes often, this stage should not be mixed into the repeat-cycle time.

Repeat bending is the recurring cycle after the job has reached a stable running condition.

Blocking movement includes only those operations for feeding, turning, stacking, or removing parts that prevent the press or operator from moving to the next cycle.

Blocking waiting covers situations where the job is ready to continue but a required resource is unavailable: the operator is busy at another machine, the tooling is being used elsewhere, the blank is not at the press, or the finished-parts area has not been cleared.

Once these components are recorded separately, it becomes clear whether the area is constrained by bending itself or by the time around it.

The shorter the batch, the more strongly changeover affects the result

Suppose two jobs require the same setup, but one contains many repeating parts while the other contains only a few. In the second case, the same preparation time is spread across fewer accepted parts.

That leads to an important conclusion: with short batches, making a single working stroke faster does not always create the largest improvement. First look at how often the tooling configuration changes during a shift and how much calendar time the machine is not bending because of those transitions.

This is why equipment and tooling manufacturers separately develop quick-clamping systems, automatic tooling seating and positioning, and automatic tool changing. TRUMPF explicitly presents automatic setup for each new program as particularly useful for small batches. WILA links fast clamping, automatic seating, and positioning assistance with time savings during tool changes.

But an automatic tool changer is not the mandatory answer for every area. First measure how much of the recurring loss is actually caused by manual tooling changes. If the main idle time comes from unavailable operators or blanks, automating tooling alone will not remove the primary constraint.

Tooling must be not only “owned” but available at the required moment

In a short-batch flow, tooling becomes a production resource in its own right.

If the required segments are stored far from the machine, mixed with other tooling, or already installed on another press, the company formally owns the set but the job still waits. WILA recommends keeping tooling near the press and making it easy to identify specifically to reduce time and errors during changeovers. Bystronic separately describes local tooling libraries near machines and duplication of frequently used sets as ways to avoid wasting time on searching and transport.

That does not mean every tool set should be duplicated. An expensive special tool used only occasionally may rationally remain a shared resource. Lower-cost segments that are needed frequently, however, can create a disproportionately large queue if they constantly move between workstations.

That is why a time study should record separately:

• searching for the required tooling;
• moving the tooling;
• waiting for it to become free on another machine;
• actual installation and positioning;
• time spent correcting an error if the wrong segment or configuration was selected.

This makes it possible to see where better organization is enough and where investment in another tooling system is genuinely justified.

The operator is also a resource with a calendar of their own

The press may be free, the program ready, and the tooling installed, but the job still cannot start without a person if the configuration requires manual feeding, support, turning, or inspection of the part.

This becomes especially visible when one operator services several machines. Such an arrangement can work well if the automatic cycles leave enough time for the person to move between presses. But if both machines need loading, first-part inspection, or difficult manual positioning at the same time, one of them will start waiting.

That is why “two presses per operator” cannot automatically be counted as two independent resources. For each work family, you need to understand:

• how much active operator time one part requires;
• which actions are tied to the moment the machine cycle ends;
• what can happen in parallel during an automatic machine stroke;
• where part mass, size, or complexity actually requires a second person;
• how much time the operator spends away from the press on tooling, blanks, containers, or documents.

In its material on increasing bending capacity, Bystronic separately notes that automation can shift the operator's role from directly servicing every cycle toward managing material flows and several cells. This is a useful way to think about the resource: count not whether a person is “present during the shift,” but the minutes when a specific machine cannot continue without that person.

A queue can grow even when average utilization looks acceptable

A monthly average utilization percentage hides the structure of the flow.

Short batches arrive unevenly. Some jobs are urgent. Some require the same tooling. Others need a longer first-part verification. If several such jobs compete at the same time for one press, one operator, or one tooling set, waiting time can grow much faster than the monthly average suggests.

For that reason, it is useful to look not only at total machine hours but also at the actual queue:

• how many batches were waiting to start bending;
• how much time passed between blanks becoming ready and the first bend;
• how often an urgent job interrupted an already configured run;
• how many repeated changeovers appeared because priorities changed;
• which resource was most often the reason for waiting.

If an urgent batch regularly forces the team to remove tooling, make a few parts, and then restore the previous configuration, its cost to the area is larger than its own bending time. It also adds two transitions between jobs and may delay other orders.

There is no universal “ideal utilization” that solves this problem. It becomes visible only in the real sequence of work.

A blank that “is in the shop” is not necessarily ready for bending

Some press-brake idle time is created outside the press itself.

The laser has already cut the parts, but they have not been sorted by batch. The cart with the required order is parked in another area. There is no space beside the press for finished parts. Large bent parts have accumulated in the aisle and interfere with safe movement of the next batch. The information system says the order is “ready,” but physically the operator still cannot begin.

The model should therefore separate:

• material feeding, movement, and unloading time that blocks the press or operator;
• movement that another worker can perform in parallel or at another time;
• waiting because a batch has not been physically prepared;
• waiting because the finished-parts area is full.

This classification matters more than the number of meters between the laser and the press. A short route can be badly organized, while a longer route can run predictably when buffers, carts, and replenishment rules are clear.

Move preparation work out of the press calendar where the resources allow it

One of the simplest ways to increase useful machine time is to perform work that does not require the press while the press is running.

For example, program preparation should not automatically consume machine time if it can be done offline. TRUMPF explicitly describes a scenario in which programming is performed from an office while the press continues operating. The same logic can apply to preparing documents, checking the required tooling in advance, placing the next batch in a buffer, and preparing containers for finished parts.

But it is important not to create imaginary savings. If the same person must bend the current batch and prepare the next one, the tasks do not become parallel merely because they do not physically require the same machine.

Every preparation activity therefore has to be assigned to a specific resource: press, operator, programmer, tooling, cart, crane, or staging area. Only then can you see what can actually leave the critical path.

A practical time study should explain losses, not just produce one percentage

For an initial analysis, use a representative period that includes typical short batches, repeating jobs, and several more difficult cases.

For each batch, record:

You can then calculate the result over the selected calendar horizon:

`Throughput = number of accepted parts / calendar time of the bending area`

With a mixed product range, a simple part count is not always sufficient because parts differ in complexity. In that case, it can be useful to also track completed batches, machine hours by part family, or another internal work equivalent. The key is not to mix different metrics as though one simple part were equal to one complex part.

If the area has several presses, analyze their calendars separately. Simply adding all hours can hide the fact that one machine is overloaded with the relevant product mix while another is free but lacks the required tooling, bending length, force, or operator.

Start improvements with the largest recurring loss

After the time study, the possible actions become much more specific.

If a large share of time disappears into changing and searching for tooling, first examine storage organization, quick-change systems, and standard stations, and only then automatic tool changing.

If the machine waits for a program, offline preparation and job-readiness discipline may be the stronger step.

If the operator repeatedly leaves the press to fetch material, the problem may be replenishment, buffers, or task allocation.

If the queue is created by several jobs that all require one special tool set, the constraint is tooling rather than tonnage or ram speed.

If, after organizational changes, repeat bending time itself consistently occupies the critical part of the calendar, only then does it make sense to compare a faster machine, an additional press, or an automated cell.

That is the purpose of the model: not to prove that the area is “loaded to the correct percentage,” but to identify which time component is actually constraining the output of accepted parts.

What to prepare for a practical discussion about the bending area

You do not need a perfect data set to estimate real throughput. Take a representative period and collect, batch by batch: part quantity, setup/changeover time, first-part time, repeat-cycle time, waiting and its causes, available tooling, and the operator's actual involvement.