Last spring we burned a 1.2 mm wide cooling groove, 12 mm deep, into a hardened A2 tool steel insert. Forty-eight grooves across the order. My plan was one graphite electrode for the whole run, which is the plan that looks good on a quote and terrible in the inspection room.
Groove one measured 1.200 mm on the CMM. So did groove ten. Groove twenty-four was still inside the plus or minus 0.015 mm band. By groove forty I was standing at the machine with a micrometer after every burn. Groove forty-eight came in at 1.160 mm at the root. Forty microns undersize, on a feature with a thirty micron band. We scrapped the insert.
Wear is not uniform, and that is the whole problem
Everyone knows EDM electrodes wear. The number a graphite supplier hands you is a volumetric wear ratio, and ours was around 0.4 percent, which sounds like nothing. But that is an average over the entire electrode, and an average is exactly the wrong statistic here.
Wear concentrates at the leading corner of the burn and on the end face of a deep, narrow feature, where the dielectric cannot flush properly and the discharge keeps re-striking the same edge. In a 12 mm deep, 1.2 mm wide groove the flushing is bad by definition. There is nowhere for the debris to go.
I measured the electrode after the run. The bottom 3 mm of the blade had lost 0.020 mm per side. The top 3 mm had lost almost nothing. The electrode did not shrink. It tapered, and it tapered in the one direction that matters.
That is why the orbit could not save us. We cut the electrode 0.15 mm undersize per side and orbited out to size, which is a fixed offset. When the blade is thinner at the bottom, the groove is narrower at the bottom. No orbital offset puts material back.
The number I should have been watching
Just under one micron per groove. That is the wear rate on groove width across the run, and it is invisible on any single part. If you inspect every tenth groove, the step between checks is eight microns, still comfortably inside a thirty micron band. Nothing looks wrong until it is forty microns wrong.
The second trap is that the last groove is the one the customer measures. We had been checking groove one and groove ten. Those are the two best grooves on the whole insert.
What we changed
Three electrodes for forty-eight grooves instead of one, and the finishing pass always on a fresh electrode. Roughing is what wears a blade, so roughing does not get to do the finish.
We now measure the electrode, not just the part. Same CMM, same fixture, every twelve grooves: blade thickness at the top, at mid-depth and at the bottom. Three numbers, two minutes. When the bottom number drifts past 0.03 mm, the electrode comes off the machine. That is the trigger, and it is a number an operator can read without asking me.
The electrode drawing got a wear allowance built in, so the toolroom cuts to the worn midpoint instead of nominal. The router now names which electrode goes in which groove range. It is boring paperwork, and it is the difference between one scrapped insert and a rework queue.
What it cost us
One insert, already rough-machined on the outside and heat treated, scrapped after the burn. Two days of machine time, a new electrode, and a second heat-treat cycle. Call it a low four-figure hit in US dollars on a job we had quoted at a healthy margin. Not fatal. Entirely avoidable.
Welding and re-burning the groove was the option we rejected. It leaves a heat-affected zone exactly where the sealing surface is, and on a mold that runs a few hundred thousand shots that is a warranty claim waiting to happen. Cheaper today, expensive in eighteen months.
If you are buying EDM work
Ask three questions. How many electrodes are on the router for this job? How do you verify electrode wear, and at what interval? And which cavity do you check, the first or the last?
A shop that answers one electrode, we check the first part is telling you the last part is unverified. A shop that can name its wear trigger and its check interval has scrapped an insert before and did not enjoy it.
We burn deep features in hardened tool steel most weeks. The rules are the same as on the milling side: define the feature that matters, pick the instrument that can actually see it, and check the part most likely to be wrong. Our sinker EDM page covers the deep-rib work, and through-profiles with sharp internal corners go to wire EDM. Hardened tool steel behaves differently in the tank than it does on a mill, which is why the practical numbers sit on the A2 tool steel and D2 tool steel pages instead of being guessed job by job.
If your drawing has a deep, narrow feature in hard material, send it over with the tolerance on that feature and we will tell you up front how many electrodes it needs. The tolerance reference sheet is a quick way to sanity-check what you have written down, and the instant quote tool returns a budget estimate to start from. Anything that needs a firm number, our team confirms within one business day. Electrode count is one of the things we check before we commit to it.