Production cycle timer

What the stopwatch is for

A cycle on a machine is two kinds of time in turn: the time the machine cuts, and the time the machine waits for a person to unload, load and press Cycle start. A phone stopwatch gives only their sum. This tool times each step, so you can see which kind of time to shorten. A faster program does nothing for a 40-second unload, and a quicker loader does nothing for a slow toolpath.

How to time

1. Set the steps of one cycle in the order they happen. The default two steps cover most jobs: machine cuts starts at Cycle start, unload and load starts when the machine stops. Mark each step as machine or manual time.
2. Press Start at the moment the first step begins, then tap the big button each time the next step begins. After the last step the next tap starts the next cycle.
3. Tap at moments you can see or hear clearly, such as the finger pressing Cycle start or the spindle stopping. A vague start point makes each reading vary with when you happened to tap, not with the process.
4. Press Odd cycle as soon as something unusual happens, such as an insert change, a chip jam or a question from the supervisor. That cycle is then left out of the mean. You can also tick or untick cycles in the table afterwards.
5. Press Stop when you leave. A cycle the stop cuts short is not counted, and the time between stopping and starting again is not counted either.

The big button also answers to Space, Enter, Page Down and the right arrow, so a Bluetooth presentation remote that sends Page Down works as a wireless button. Page Up, the left arrow and Backspace undo the last tap. Where the browser allows it, the screen stays on while timing.

On a phone, press Full screen to fill the screen, and install the tool as an app on the home screen: on Android with the install button that appears, on an iPhone in Safari with Share, then Add to Home Screen. The app opens with no signal, which is common at a machine in a workshop.

Work done while the machine cuts

Do not time deburring or gauging the previous part as a step if the operator does it while the machine cuts. As long as it ends before the machine stops, it adds nothing to the cycle. If the machine stands waiting for it, that wait already shows in the manual step, which is exactly what you want to see. The same holds for two vices or two pallets: loading one while the machine cuts the other overlaps with the cycle and adds no time.

Several machines and stations

Each machine or station is one study. Press + Add machine and enter its name: the new study takes the steps and part of the one before, so you can walk to the next machine and start timing. Pick a machine from the list at the top to switch. A study that is timing keeps running while you switch, so you can follow two machines run by one operator, as long as you tap the right one.

Part 4 compares the time per part of every station, taking it that each part passes every station once. Parts leave the line only as fast as the slowest station lets them through, because every part has to wait its turn there. Speeding up any other station only builds a pile in front of the slow one. That slowest station is the bottleneck, and the first place to work on. Where two machines do the same operation side by side, time one of them and enter 2 in Machines on this operation: the station then passes parts twice as fast.

Where one operator runs several machines, add a step Waiting for operator (manual) between the machine cutting and the unload and load. Tap when the machine stops and again when the operator reaches it. The step shows how long the machine stands waiting. If it is long, giving the operator another machine gains no parts: the time won on the new machine is lost waiting at the others.

Download Excel gives one file. Its first sheet names the bottleneck, compares every station in a table, and draws each station's time per part as a column, machine time with manual time standing on it: the tallest column, marked ▼ bottleneck in red, is the bottleneck, its two colours show which kind of time to work on, and a column above the dashed takt line cannot keep up with the demand. A station left out of the comparison is listed below the others in the table and is not drawn, since it is not part of the line. Then comes a sheet per machine with its setup, its summary, the time per part and parts per hour, the check against the takt time, the time of each step, and every cycle with its start and end. The numbers in the file are formulas, not fixed values: each machine's sheet works its means out from its table of cycles, the first sheet reads the machines' sheets, and the chart reads the first sheet. So the file can be used to try a change. Type a shorter time into a step of the cycle table, change a cycle's Used cell to leave it out, or change the machines on an operation or the parts needed per shift, and the table, the bottleneck and the chart follow. The studies stay on the phone that timed them. Where several people time different stations, each downloads the Excel file and the files are put together afterwards.

Examples

Each link opens a set of studies already timed, so every part of the tool shows: the tables, the charts, the comparison of stations and the Excel file. Your own studies are kept while you look, and Back to my studies brings them back.

• A line of 3 stations: flange F-120 goes through two lathes at OP10, a mill at OP20 and a deburr bench at OP30. The takt time is 450 min ÷ 500 parts = 54 s. The bottleneck is the mill at 63.6 s per part, so the line makes about 57 parts an hour, short of the 67 needed. Yet the mill cuts for only 49.6 s, under the takt time; the other 14 s are unloading, cleaning and loading. That is the part to fix, not the program: for example, a second fixture so the next part is loaded while the machine cuts. One lathe takes 100 s per part, and with only one, OP10 would be the bottleneck instead; the two together pass a part every 50 s. Cycle 7 on the mill had an insert change and is left out.
• One operator, two machines: one operator runs MC-04, which cuts for 75 s, and MC-05, which cuts for only 42 s. MC-05's Waiting for operator step averages 27 s a cycle, because when MC-05 stops the operator is usually busy unloading and loading MC-04. A faster program on MC-05 would only make it wait longer; MC-04 hardly waits at all.

Normal cycles and odd cycles

The mean cycle time uses the normal cycles only, because that is the number to compare with the program and to plan on. Odd cycles are still real losses, so the results also give the time per part over every cycle timed. When the two differ a lot, interruptions are costing more than the cycle itself, and the fix is in the interruptions, not in the program.

How many cycles to time

The "±" under the mean is the 95 % confidence interval of the mean: the range in which the long-run average very likely lies, judging from the cycles timed so far. It comes from the spread of the cycles (the standard deviation) and narrows with the square root of their number, so four times as many cycles halve it. Keep timing until the ± is smaller than the accuracy you need. The interval assumes the cycles vary at random around one average; if they drift steadily longer, for example, timing more cycles will not settle the number, and the drift is the thing to find.

Against the machine's own cycle time

The machine's step should match the CYCLE TIME the control shows on its position screen. Time cycles that run straight through, with rapid and feed override at 100 %, not the first part run block by block, or the numbers will not match the program. How to turn a measured cycle time into a factor for estimating the next program is in the article Estimating cycle time from an NC program (in Thai).

Takt time and parts per hour

Takt time is the working time divided by the parts needed: the time one part may take for the line to keep up with demand. If the measured time per part is longer than the takt time, one machine cannot keep up, and the number of machines needed is the time per part divided by the takt time, rounded up. Look at the longest cycles as well as the mean. With a mean just under the takt time and many cycles over it, the line falls behind every time a long cycle comes, and needs parts in stock between machines to absorb it.

Parts per hour is 3600 divided by the time per part. It assumes the machine never stops, so a shift gives fewer parts than this once breaks, changeovers and odd cycles are counted.