Intelligent and adaptive controller for a metal cutting process.

Belgiu, George ; Ruset, Vasile ; Pamintas, Eugen 等

1. INTRODUCTION

In any research for prediction, control and optimization of the metal cutting process, digital models are essential. Particularly essential are digital models for cutting force, cutting temperature, tool wear, machine tool vibration, tool breakage, chatter etc. Modern Computer Aided Manufacturing systems can offer some artificial intelligence features in the course of technologic design, to assist in the process of selecting tools, inserts, holders and also the cutting parameters ([a.sub.p]--depth of cut, f--feed and v--cutting speed). The common practice: after the technologic design was finished, all technological data -- geometric path, necessary tools and cutting parameters for the cutting process -- are transmitted to the machine tool's controller. Constantly, each time there is an important variance between the theoretical data -- NC program -- and the true metal cutting process. As we can see in figure 1, the final part quality dependency in metal cutting process is less than 20% for the CAM system.

The primary practice to obtain performance and a high quality parts, remain to fine tune the cutting process at workshop level.

[FIGURE 1 OMITTED]

2. CRITICAL OVERVIEW

Numerous expert systems were created researches worldwide, and quite a lot of scientific literature was conceived to solve the problem (Childs et al., 2000; Marusich, 2001; Marusich et al., 2009; More et al., 2006; Yallese et al., 2005).

[FIGURE 2 OMITTED]

As we can see in figure 2, the actual state for the CAM systems in conjunction with the expert systems has a technological limit. There is no feedback from the real cutting process. Also, some adaptive Computer Numerically Controls were created, but these equipments did not worked at industrial expectations, frequently because the large number of sensors (and parameters) needed to be controlled.

3. RESEARCH COURSE. TECHNICAL SOLUTION

For turning, drilling and milling (the preponderance of metal cutting processes) we selected a number of basic parameters [T.sub.i], [D.sub.i], [M.sub.i]. For turning we have Ti basic parameters:

* [T.sub.1]--high wear clear of the rake surface;

* [T.sub.2]--deformation of cutting edge;

* [T.sub.3]--galling of cutting edge;

* [T.sub.4]--cracks perpendicular to the cutting edge;

* [T.sub.5]--spalling of the cutting edge;

* [T.sub.6]--fracture of the indexable insert;

* [T.sub.7]--long spiral chips;

* [T.sub.8]--vibrations in the process.

In the same way, for drilling cutting process we defined these basic parameters [D.sub.i]:

* [D.sub.1]--drill point damaged;

* [D.sub.2]--wear on outside diameter;

* [D.sub.3]--resulting hole too large;

* [D.sub.4]--chips stuck in the flutes;

* [D.sub.5]--spalling of the cutting edges;

* [D.sub.6]--resulting hole non-circular;

* [D.sub.7]--short tool life;

* [D.sub.8]--vibrations in the system.

Finally, for the milling process we defined these eight Mi basic parameters:

* ([M.sub.1] to [M.sub.8])--similar as ([T.sub.1] to [T.sub.8]), except [M.sub.7]:

* [M.sub.7]--poor surface quality

For these troubleshooting parameters, we can have more than one corrective measure (basic measures):

* [C.sub.1]--change cutting tool geometry and holder;

* [C.sub.2]--increase supply of lubricant;

* [C.sub.3]--increase / decrease feed [f];

* [C.sub.4]--increase / decrease cutting speed [[v.sub.c]];

* [C.sub.5]--increase / decrease cutting depth [[a.sub.p]];

* [C.sub.6]--change carbide type.

Basically, the solution S for this problem is a "simple function" like in equation 1, a correlation between basic- process parameters and corrective measures:

S = f([T.sub.i], [D.sub.i], [M.sub.i]|[C.sub.i]) (1)

For signal acquisition we consider as constrains: force, torque, spindle speed, acoustic noise, displacement in turret, temperature in tool tip, and acceleration of carriage. For signal processing we use: Fourier transfor-mation, wavelet transformation, statistics (mean, variance, skew) and wave shape profile (peak, slope, envelope).

The cyclic activities of an intelligent and adaptive controller are shown in figure 3: sensing, processing, recognition and decision. In essence, we have two groups of monitoring: direct and indirect signals. Direct monitoring refers to cutting force, tool and workpiece vibration, tool wear, tool chipping, and most important for safety--tool breakage. Indirect monitoring refers to chip control, actual depth of cut, surface quality and dimensional error.

The controller is monitoring the cutting conditions as is shown in figure 4. For every parameter ([T.sub.i], [D.sub.i], [M.sub.i]) we have time dependent threshold. The intelligent and adaptive controller is thanking the appropriate decision, by computing intensive tasks in cutting processes (in terms of energy, simulations and dynamic interactions. Eight basic parameters are permanently computed, depending of cutting process: turning / drilling /milling.

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

At this moment, any microprocessor of a CNC controller is not capable to solve the amount of data in real time. For simulation we used the parallelization method.

4. CONCLUSIONS. FURTHER RESEARCH

In this paper we attempted to capture in real time the decision taken by the manufacturing engineer / operator of a CNC machine tool during the metal cutting process.

An expert system was created, however is annexed to the metal cutting process, and at this moment is useful only to improve the NC programs--to better design the cutting process.

At the moment, the expert system is running on separate hardware machine (PC host, Linux operating system), separated from the CNC controller. In parallel with the features of the expert system, a further research is to include an expert system directly on the CNC controller, to steer completely the execution of a mechanical part.

Also, further developments of the expert system must consider a better distinction within the type of cutting process (turning, milling or drilling).

5. REFERENCES

Childs, T.; Maekawa, K.; Obikawa, T. & Yamane, Y. (2000). Metal Machining. Theory and Applications, John Wiley & Sons Inc., ISBN 0 470 39245 2, New York

Marusich, T.D. (2001). Effects of Friction and Cutting Speed on Cutting Force, Proceedings of ASME Congress 2001, November 11-16, 2001 New York, NY. Available from: http://www.thirdwavesy s.com/news/published_papers.htm. Accessed: 2009-05-23

Marusich, T.D.; Stephenson, D. A.; Usui, S. & Lankalapalli, S. (2009). Modeling Capabilities for Part Distortion Management for Machined Components. Available from: http://www.thirdwavesys.com/news/published_papers.htm. Accessed: 2009-05-23

More, A.S.; Jiang, W.; Brown, W.D. & Malshe, A.P. (2006). Tool wear and machining performance of cBN-TiN coated carbide inserts and PCBN compact inserts in turning AISI 4340 hardened steel, Journal of Materials Processing Technology, vol. 180, pp. 253-262

Yallese, M.; Rigal, J.F.; Chaoui, K. & Boulanouar, L. (2005). The effects of cutting conditions on mixed ceramic and cubic boron nitride tool wear and on surface roughness during machining of X200Cr12 Steel (60HRC), Proceedings of the Institution of Mechanical Engineers, Part B. Journal of Engineering Manufacture, vol. 219, pp. 35-55