Joint use of dynamical classifiers and ambiguity plane features

Multirate Coupled Hidden Markov Models and Their Application to Machining Tool-Wear Classification

“…Wear inherently progresses over multiple time scales, ranging from distinct phases of cutting within a cutting pass (such as entry to the piece, bulk cutting, and exit from the piece) to transients and other shortterm phenomena (e.g., due to material removal) occurring at the millisecond level. In particular, acoustic vibrations from titanium milling experience noisy/quiet transient periods that can last up to thirty seconds [40]. Such temporal information can be a good indicator of the wear amount, but existing (static or dynamic) monitoring systems do not utilize it.…”

“…Titanium is chemically reactive, and tends to weld to the cutting edges during cutting, and form a so-called built-up edge in the milling jargon [40]. As the built-up edge volume exceeds a threshold, the bonding forces are overcome by the cutting forces, leading to the chipping of pieces off the primary cutting edges, and premature tool failure.…”

“…Much of the previous work in milling tool-wear prediction has focused on steel with few results reported on titanium [40], [23], in part because titanium poses new challenges. Titanium is chemically reactive, and tends to weld to the cutting edges during cutting, and form a so-called built-up edge in the milling jargon [40].…”

“…Tools used in our milling task are relatively small (1/2 in diameter), and cutting forces, current, and power are not very sensitive to track wear on such small size tools [14], [24]. We use acoustic vibrations instead, as they have been successfully used in similar tasks [41], [1], [23], [40].…”

“…Each welding and release cycle can last from one second to thirty seconds, and the quiet periods are characterized by reduced cutting forces, vibrations, and the rate of wear progression. It has been hypothesized that the primary reason for why the existing systems particularly fail when applied to titanium is that they do not model such long-term characteristics [40]. Among other difficulties associated with titanium are its low thermal conductivity, increasing temperature at the cutter-workpiece interface, and the heterogeneous structure of titanium, containing random hard spots [20].…”