Direct answer: In a robot machine tending cell the arm is idle for most of the cycle. Machine tending cycle time is set by machine processing, the door and chuck sequence, and how reliably parts are presented — not by the robot’s top speed. Size the cell from the machine cycle first, then decide whether one robot serves one machine or several. A faster arm only shortens the handling portion.
Who this is for: This guide is written for production and automation engineers sizing a robot machine tending cell, and for buyers who have been quoted a cycle time and need to know what it actually depends on.
Scope: It covers where the time in a tending cycle really goes, what the robot controls and what it does not, and how to measure the result honestly. It does not cover cutting parameter selection, machine tool capability, or fixture design inside the machine.
Where machine tending cycle time actually goes
Machine tending cycle time only makes sense once the cycle is broken into its real segments, and the picture changes quickly when it is. There is a pick from the staging position, a move to the machine, a load and clamp handover, a withdrawal, a period during which the machine processes the part, an unload, and a return to staging. In the reference footage for this guide the robot picks a ring part from a staged tray, places it into the lathe chuck, and then withdraws from the working area while the machine cuts.
EVST breaks a tending quotation into exactly those segments. Of those segments, the robot controls the moves and the handover. It does not control the processing period, and on most turning, milling, and grinding work the processing period is the largest single block. That is why arm speed has a bounded effect on the cell’s output, and why quoting a cell on arm speed alone tells you very little.
It also explains a result that surprises people the first time they measure it: making the robot faster can leave the cycle unchanged, because the arm simply arrives at its waiting position sooner. The gain shows up only if the handling segment was on the critical path to begin with. Understanding what robot machine tending is and how it works is largely understanding which segment is critical.
The handover is the constraint, not the arm
The handover — approach, insert, clamp, confirm, release, retract — is where the real variability lives. Chuck or vice closing time, door opening and closing, coolant and chip clearance, and the confirmation signal that says the part is seated all sit in this window. Each of those is a machine or fixture property, not a robot property.
EVST treats the handover as the specification, not the arm. This is also where reliability is won or lost. A handover that works ninety-nine times in a hundred produces an unattended cell that has to be attended. Positive part-present confirmation, a defined recovery for a failed seat, and a chip management strategy are worth more to real output than a faster move profile.
The signals themselves need to be agreed with the machine tool builder early. What the control can expose — cycle end, door state, chuck state, alarm state — determines how tightly the cell can be sequenced. According to IEC 60204-1:2016, general requirements apply to the electrical equipment of machines, which is the framework in which that interface is specified and wired rather than improvised.
Part presentation decides whether the cell runs unattended
A tending cell only runs unattended for as long as it can find the next part in a known position. Trays and pallets give a known position at the cost of an upstream sorting step. Bins and conveyors reduce that step but demand sensing and add variation. The choice is a material-flow decision with an automation consequence, and it is usually made too late.
Presentation also decides gripper design. A part that always arrives in the same orientation can use a simple two-jaw gripper; a part that arrives loosely oriented needs either a re-grip station, a compliant approach, or vision. Each of those is a real addition to the handling segment, which is the one segment the robot actually controls.
The finished part needs the same care. In the reference footage the robot returns the machined part to the staged tray, which is the simplest reliable pattern: the same position family for in and out, so a single teaching scheme covers both.
Decision table: what evidence justifies which tending configuration
Every row here can be settled with data you already have or can measure in an afternoon. None of it requires the robot to be selected first.
| Evidence from the machine and the part | Configuration it justifies | Evidence you still owe |
|---|---|---|
| Long processing time, one machine | One robot serving several machines | Worst-case waiting when two machines finish together, walking or rail travel time, alarm handling |
| Short processing time, one machine | One robot dedicated to that machine, focus on handover | Handover time breakdown, door and chuck sequence, confirmation signal reliability |
| Parts arrive loosely oriented | Sensing or a re-grip station before load | Added handling time, failure mode when orientation is not found, recovery procedure |
| Unattended running required over a shift | Staging capacity sized to the shift, chip and coolant management | Buffer count actually needed, chip clearing interval, what stops the cell safely |
| Mixed parts on the same machine | Managed program and gripper set with positive identification | Changeover procedure and time, wrong-program prevention, first-off rule |
One robot, several machines
Serving several machines with one robot is the standard way to convert a long processing time into utilisation. It also introduces a new failure mode: contention. When two machines finish inside the same window, one of them waits, and the cell’s effective output is set by how often that happens rather than by the average.
EVST models contention explicitly before a multi-machine layout is fixed. The honest way to evaluate this is to model the worst case, not the average, and to include travel between machines, alarm handling, and the operator interventions that will actually occur. If the machines have similar cycle times and start together, contention is frequent; if their cycles differ, it is rarer. Neither case is a reason to avoid multi-machine tending, but both should be stated before the layout is fixed.
Layout follows from that decision. A linear arrangement on a rail, a central robot serving machines around it, and a mobile arrangement each trade travel time against floor area and safeguarding complexity, and each belongs in the comparison rather than being assumed.
Safeguarding an interlocked machine and robot
A tending cell is two machines sharing a space and a sequence. The risk assessment has to cover the combination, including what happens when the machine door opens while the robot is in motion, and what state each machine is left in when a protective stop occurs. According to ISO 12100:2010, the general principles for design cover risk assessment and risk reduction, and according to ISO 10218-2:2025, safety requirements apply to the industrial robot application and cell.
The practical outcome of that assessment is usually a defined interlock set and a defined safe state for each of the two controls, written down and tested at acceptance rather than discovered during a fault. Recovery — how an operator clears a part stuck in the chuck without defeating a guard — belongs in the same document.
This is also the part of the specification that survives longest. Cells get reprogrammed and re-tooled; the safeguarding concept and the interlock contract usually stay.
Measuring the result honestly
Once a cell runs, the temptation is to report machine tending cycle time as a single number. A more useful record separates handling time, machine processing time, and waiting, and reports them over a period long enough to include the interruptions that really happen. According to ISO 22400-2:2014, key performance indicators for manufacturing operations management are defined, which gives a common vocabulary for that reporting instead of a locally invented one.
For the robot itself, according to ISO 9283:1998, performance criteria and related test methods for manipulating industrial robots include the conditions under which pose accuracy and repeatability figures are established. It is worth knowing that those figures are measured under defined load and speed conditions, because a datasheet number quoted outside those conditions is not a promise about your part.
This guide asserts no cycle time, output, or utilisation figure for any cell. The footage behind it shows a pick, a load into a chuck, a withdrawal, and an unload; it does not demonstrate a rate, and none is claimed here. If you are comparing automatic welding system solutions and tending cells across a plant, that distinction is what keeps the comparison meaningful.
What to send for a tending study
A useful study starts from the machine, not the robot. Send the part drawing with mass and gripping surfaces, the machine’s processing time and its door and chuck sequence, the way parts arrive today, the number of machines you want covered, and the interlock signals the machine control can provide. With that, the handling segment can be estimated and the waiting segment can be calculated rather than guessed.
What comes back should tell you which segment is critical and what would have to change to move it. If the answer is that the machine dominates, the honest recommendation may be a simpler robot and better staging rather than a faster arm.
Frequently asked questions
Will a faster robot shorten our machine tending cycle time?
Only if the handling segment is on the critical path. When the machine’s processing time dominates, a faster arm arrives at its waiting position sooner and the cycle is unchanged. Measure the segments before paying for speed.
Can one robot serve several machines?
Often yes, and it is the usual way to use a long processing time. The thing to evaluate is contention: how often two machines finish in the same window, and what the worst case costs. Model the worst case, not the average.
What most often stops an unattended tending cell?
Part presentation and chip or coolant management, followed by handover confirmation failures. None of those are robot speed problems, and all of them are cheaper to solve at design time than after installation.
How should cycle time be reported at acceptance?
Separately by segment — handling, processing, waiting — over a period long enough to include real interruptions. ISO 22400-2:2014 defines key performance indicators for manufacturing operations management and gives a shared vocabulary for that report.
Project inputs for an application review
Send the following and the tending cycle can be reviewed against your machine rather than a generic figure:
- the part drawing with mass, gripping surfaces, and orientation requirements
- the machine cycle: processing time, door and chuck sequence times, and any in-cycle gauging
- the current staging method — tray, bin, conveyor, or pallet — and how parts arrive
- the number of machines you want one robot to serve, with their layout
- the interlock signals your machine control can already provide
Send the part drawing with mass and gripping surfaces, the machine processing time with its door and chuck sequence, how parts arrive today, the number of machines one robot should serve, and the interlock signals your machine control can provide. That set is enough to break the cycle into segments and say which one actually limits output. Related reading: collaborative robot options.