CHAPTER 6 ON-LINE TOOL WEAR COMPENSATION AND ADAPTIVE CONTROL

Size: px

Start display at page:

Download "CHAPTER 6 ON-LINE TOOL WEAR COMPENSATION AND ADAPTIVE CONTROL"

Abel Booth
5 years ago
Views:

1 98 CHAPTER 6 ON-LINE TOOL WEAR COMPENSATION AND ADAPTIVE CONTROL 6.1 INTRODUCTION There is lot of potential for improving the performance of machine tools. In order to improve the performance of machine tools, a better understanding of the process is necessary. Since machining processes are highly non-linear and difficult to predict, real-time, on-line process monitoring and control is a feasible alternative. Tool wear affects product quality in the form of poor dimensional accuracy and surface finish. Hongli Gao et al (2006) have suggested a selfadjusting tool wear monitoring system. They have improved the classifying accuracy of tool wear and solved problems related to the high cost of tool condition monitoring system. In this work an attempt has been made to develop a real-time monitoring and real-time adaptive control system for HSM. Earlier attempts have been made to regulate cutting parameters using PID- controllers by Ziegler and Nichlos (1942) and by Centner and Idelsohn (1964). Tlusty and Elbestawi (1977) and Tomizuka et al (1983) attempted fixed gain adaptive and Model Reference Adaptive Control (MRAC) for process control respectively. Milner (1973) has categorized adaptive control system into two major groups, i.e. Geometrical Adaptive Control (GAC) and Technological Adaptive Control (TAC). TAC is further divided into two sub-

2 99 categories: Adaptive Control with Constrains (ACC) and Adaptive Control for Optimisation (ACO). 6.2 ADAPTIVE CONTROL Adaptive control (AC) is the next level in CNC systems. The current practice of setting speed, feed, and depth of cut is based on operator/ programmer skills. However based on close observation of process parameters and knowledge of work piece, tool material and other factors, online adjustments of operating parameters is possible, subject to machining condition to optimize the overall system performance. An adaptive control is a controller with adjustable parameters and a mechanism for adjusting the parameters. The adaptive control systems are generally closed loop systems. The AC systems currently used take control of more than one parameter. In the present CNC systems the part programmer decides the operating parameters and the real time process variables are not accounted. A conservative approach to save tool life and part quality is generally used. This approach results in slowing down the production rate and overall system performance. With the availability of dedicated computers, hardware and programming software the development of adaptive control has taken new dimension. In spite of much research in this area, Galip Ulsoy and Yorem Koren, (1989) say that commercial applications would take a longtime Development of Process Monitoring and Adaptive Control System The development of process monitoring and adaptive control is discussed in the module 3 and module 4 as follows:

3 100 Module 3 - To design an self-adjusting on-line tool wear compensation system in HSM. Module 4 - To develop a system for adaptive control based on constraints (ACC) for constant cutting force and improved surface finish. 6.3 MODULE -3 DEVELOPMENT OF SELF-ADJUSTING ON- LINE TOOL WEAR COMPENSATION SYSTEM In order to accomplish the objective of developing an on-line tool wear compensation system, the following steps are followed: Self-adjusting on-line tool wear compensation system is developed and tested A process capability study is conducted for verification of new process. The algorithm for tool wear compensation is given in Figure 6.1. The basic approach used in this work is pre-process characterization of tool wear based on experimental data and artificial neural network and a real-time monitoring of milling process to enable real-time adaptive control of the high speed milling process. The set-up will make realtime measurements of acoustic emission signals and correlate with tool wear for a given cutting condition. The above data are used as real-time input data to an algorithm that will compensate with work offset after the predefined value of tool wear is attained. In vertical CNC machining center, machine tool (cutting tool-primary datum) datum and work datum are used. The NC codes are generated based on the above datum. In the current shop floor practice,

4 101 Figure 6.1 Algorithm for On-Line Tool Wear Compensation System shifting the machine datum in the NC program is carried out according to measured tool wear. However in this work, an attempt is made to indirectly monitor tool wear using acoustic emission sensor and compensate by workoffset method. The end-milling cutter is subjected to both side and end flank wear. There is a predominant side flank wear when compared with end flank wear. However in this study, end flank wear is addressed and work compensation is introduced in the Z-axis. The end flank wear effects typical components like

5 102 O ring groove depths in hydraulic circuits, depths in spigot and recess applications, dimensions of snap fit joints in injection molding dies and assembly of monolithic aircraft structures. Figure 6.2 represents the typical O ring groove taken for the study. Figure 6.2 Typical O ring groove Experimental Set-up The system has been designed to measure acoustic emission signals during machining. The process variables such as cutting speed, feed rate and depth of cut are fed as input to an external computer. The cutting parameters together with the AE signals is used as input data to an algorithm which decides the shifting of work datum during progressive tool wear with the help of a stepper motor drive. Figure 6.3 shows the experimental set-up which comprises of the following: 1) Vertical Machining center 2) Work piece and work holding device 3) Stepper motor drive and driver circuit 4) AE sensor, filter and amplification circuit 5) USB 6009 data acquisition card (DAQ)

103 Work piece Stepper motor AE sensor DAC 6009 Slide Figure 6.

milling cutter, work piece and AE sensor were discussed in chapter 3.

this set-up and their details are discussed below. 6.3.1.

specification is used: Torque: 0.5 Nm Step angle: 5 0 (full step); 2.

6 103 Work piece Stepper motor AE sensor DAC 6009 Slide Figure 6.3 On-line tool wear compensation system The high speed milling set-up, end milling cutter, work piece and AE sensor were discussed in chapter 3. The stepper motor, driver circuit and USB 6009 DAQ card are new additions in this set-up and their details are discussed below Stepper motor and driver circuit The stepper motor of following specification is used: Torque: 0.5 Nm Step angle: 5 0 (full step); (half step) Current: 3 A Voltage: 10 V Scheme: Unipolar (4 winding) Linear displacement of slide: Half step: 7 m, full step: 14 m

7 104 The stepper motor drive setup consists of a micro-controller, driver circuit and a stepper motor. The driver circuit energizes the stepper motor by giving a required voltage to the phase windings. The circuit consists of a Darlington pair of transistor for each phase winding, in order to drive the stepper motor. Diodes are provided for unidirectional current flow. Triggering signals are generated from the micro-controller unit. In full step mode only one phase winding is energized, whereas in half step mode two-phase winding is energized at a time. The required movement of the slide is achieved by the half step mode. Figure 6.4 shows the block diagram of the stepper motor driver circuit. Figure 6.4 Block diagram of stepper motor driver circuit An USB 6009 is used to trigger the micro-controller. The work piece of standard size as per ISO 8699 (1989) is mounted on a vertical slide that is driven by the stepper motor.

8 Data acquisition card (DAQ) The data acquisition card forms the connectivity between the computer and the real world. The DAQ receives different types of input signals from various sensors through the different analog and digital channels available. An USB 6009 (Universal serial bus) type data acquisition card is used with 12/14 bit input resolution and can handle up to 48ks/s. It comprises of a 32 bit counter, with 12 digital I/O lines and 2 analog output LabVIEW The LabVIEW software is programmed to receive the input from the sensors and adjust the work datum during progressive tool wear. The following sequence of operations is involved: Acquiring AE signals from DAQ through analog input. Filtering of the AE signal and storing the filtered Aerms data. Development of reference table for correlating AE signals and tool wear. Comparison of acquired AE signals with values in reference table If the obtained signal matches with the value in the reference table then the system gets ready for compensation. A pulse is to be given to the stepper motor drive circuit through the digital output of DAQ card for compensation in steps of seven micrometer. The tool wear is compensated in steps of seven micrometers due to hardware limitation. Tool

9 106 wear could be compensated even with smaller step value with better hardware. Figure 6.5 shows the front panel of the LabVIEW software. Here the signals acquired from the DAQ card are first filtered and processed by means of filter module. Then the Aerms values are stored to a LabVIEW measurement file and simultaneously the waveform indicator shows the nature of the waves obtained from AE signals. The Aerms, which are obtained through LabVIEW, are then compared with the AE value in reference table that have been built up and each time the Aerms obtained is same as the values in reference table, a pulse is sent to the stepper motor for compensating the tool wear of seven microns. On-line AE data Reference value. Figure 6.5 Front panel of LabVIEW software

10 Testing of on-line tool wear compensation system In order to estimate the capability of the tool wear compensation system, a set of experiments along with process capability study is carried out. Two sets of experiments are conducted with one new tool each. In the first case the tool wear compensation is not used, instead the conventional practice of tool offset at regular interval is carried out. Table 6.1 gives the details of the experimental conditions. Table 6.1 Experimental condition Sl.No Specification 1 Cutting Speed rpm 2 Feed rate 4000 mm/min 3 Depth of cut 0.25 mm/pass 4 Work piece Aluminium LM6 5 Tool 6mm, solid TAlNi coated carbide two flute end mill 6 Oil groove depth 1007 m 7 Tolerance H 8 8 Specification limits LSL= 1000 m, USL= 1014 m 9 Sample size 60, during steady wear period of tool In the second case the tool wear compensation system is used. Both experiments wear carried out separately and for ease of comparing the results the readings are shown in Table 6.2

11 108 Table 6.2 Experimental observations S.No. Part No. Machining Time,s Aerms,V Predicted Tool Wear ( m) Measured groove depth ( m) Without compensation With compensation

12 109 S.No. Part No. Machining Time,s Aerms,V Predicted Tool Wear ( m) Measured groove depth ( m) Without compensation With compensation

13 Process capability study Before going into the process capability study the following assumptions are made. The interpretation of capability indices such as C p and C pk rests on a base of several assumptions. 1. The process is in a state of statistical control. 2. Sufficient data are collected during the capability study to minimize the sampling error for the capability indices. 3. The data are collected over a sufficient period of time to ensure that the process conditions present during the study are representative of current and future conditions. 4. The parameter analyzed in the study follows a normal probability distribution. Otherwise, the percentages of product associated with values of C p and C pk are incorrect. To find out whether the new process has given better result, a process capability study for the processes is devised The process capability ratio (Cp) is defined in equation (2) USL LSL Cp 6σ (6.1) where, Cp - Process Capability Ratio USL - Upper Specification Limit LSL - Lower Specification Limit σ - Standard Deviation of the process Experimental observations from Table 6.2 has been taken for the process capability study Figure 6.6 (a) and (b) shows the control charts without and with tool wear compensation respectively.

14 111 Dimensions Without On-line tool wear compensation system UCL CL LCL Components (a) Dimensions With On-line tool wear compensation system UCL CL LCL Component (b) Figure 6.6 (a) Control chart (Without compensation) (b) Control chart (With compensation) The repeated trend in the control chart (a) is due to the tool offset for every ten components and in control chart (b) is due to tool wear compensation.

15 112 From the above control chart, it is observed that the component dimensions are in statistical control. This satisfies the first assumption of process capability analysis i.e., the process should be in statistical control. (a) Figure 6.7 (b) (a) Control chart (Without compensation) (b) Control chart (With compensation) From Figure 6.7(a) and (b), it is observed that the components machined without and with on-line tool wear compensation system follows normal distribution. This

16 113 satisfies the second assumption made in process capability analysis i.e., the process should follow normal distribution Comparison of results Table 6.3 shows the comparison between the two processes. Table 6.3 Comparison between the two processes Name of the statistic Three sigma limits Specification limits Without compensation system LCL= UCL= LSL= USL= With compensation system LCL= UCL= LSL= USL= Process Mean (µ) Standard Deviation (σ) Variance Process Capability Ratio (Cp) Process Capability Index (Cpk) where, LCL Lower Control Limit UCL Upper Control Limit The mean of the process with on-line tool wear compensation system has improved when compared to the mean of the process without online tool wear compensation system. The standard deviation of the process

17 114 with on-line tool wear compensation system has reduced when compared with standard deviation of the process without on-line tool wear compensation system, due to less scatter. Further, the process capability indices and process capability ratio have improved in on-line tool wear compensation system. During the experimental study, the surface roughness value (R a ) was found to be within 0.8 to 1.6 µm. 6.4 MODULE 4 DEVELOPMENT OF ACC FOR IMPROVED SURFACE FINISH In order to develop an adaptive control based on constraints (ACC) for constant cutting force and improved surface finish, the following objectives are to be satisfied: To develop and to test an ACC for constant cutting force and improved surface finish in HSM. To conduct a process capability study for verification of new process. steps: The construction of an adaptive controller involves the following i) Characterize the desired behavior of the closed loop system. ii) Determine a suitable control law with adjustable parameters. iii) Find a mechanism for adjusting parameters. iv) Implement the control law. The above steps have been used in module 4. The algorithm for ACC is given in Figure 6.8.

18 115 START MACHINING IN 3 AXIS CNC VERTICAL MILLING MACHINE CUTTING FORCE MEASUREMENT FROM NO END CONTINUE? YES INCREASE IN CUTTING SPEED BY INVERTER VOLTAGE SIGNAL FROM DAQ TO INVERTER DAQ SAMPLING CUTTING FORCES IN PC WITH LabVIEW NO CUTTING FORCE > 24 N? YES COMPARISON OF CUTTING FORCE VALUE WITH THRESHOLD VALUE Figure 6.8 Algorithm for ACC The set-up will make real-time measurements of cutting force using a cutting tool dynamometer during machining. Tool wear results in increase in cutting force and poor surface finish. Therefore maintaining a constant cutting force indirectly means, keeping the surface finish value also at desired level. Alpay Yilmaz et al (2002) have attempted to suppress chatter in machine tool by random spindle speed variation. Here again a similar attempt is made, but instead of chatter suppression the cutting force is reduced. The cutting force signal from dynamometer is used as real-time input data to an algorithm that

19 116 will vary the cutting speed by varying the spindle speed such that the resultant cutting force is maintained below the threshold value. In the vertical CNC machining center, the cutting parameters are fixed by the programmer/operator and are seldom changed. The spindle speed variations generally need a change in the NC codes. However in the VMC used, in the current work has a separate spindle controlled by frequency drive. Therefore it is easy to vary the spindle speed using a DAQ card and LabView program. The aluminium components used in typical applications like hydraulic circuits, medical equipments and aircraft components are processed in conventional milling machines and are further ground to obtain the required surface finish. The typical surface finish (Ra) values of the order of 3.2 m are achievable in conventional milling process. The components are further subjected to grinding process to reach surface finish values of 0.8 to 1.6 m. In HSM applications typical surface finish values of 0.8 to 1.2 m are obtained without grinding. However as the tool becomes blunt the surface finish is poor. It is in this context that the ACC developed will overcome the existing drawback in HSM Experimental Set-up The system has been designed to measure cutting forces during machining. The process variables such as cutting speed, feed rate and depth of cut are fed as input to an external computer. These together with the cutting force signal, is given as input data to an algorithm, which decides the change in cutting speed by varying the spindle speed. Figure 6.9 shows the schematic representation of the experimental set-up, which comprises of the following: 1. Vertical Machining center 2. High speed spindle and inverter

20 Milling dynamometer 4. USB 6009 data acquisition card (DAQ) 5. Personal computer Figure 6.9 Schematic representation of experimental set-up with ACC The high speed milling set-up, end milling cutter, work piece, cutting tool dynamometer, LabVIEW software and USB 6009 DAQ card are discussed in detail in chapter 3 and chapter 6. However the spindle speed control and RS 485 cable with Commander SE inverter are new additions and their details are discussed below Spindle speed control using ACC An integral motorized spindle of 5kW and rpm is used in the VMC. The spindle is driven using a frequency control inverter. The frequency

118 can be set manually or from the personal computer through an RS485 cable arrangement. A commander SE software is used to control the spindle speed through the computer.

21 118 can be set manually or from the personal computer through an RS485 cable arrangement. A commander SE software is used to control the spindle speed through the computer. The USB 6009 DAQ card receives the signals from the dynamometer and sends it to the LabView program. The LabVIEW is programmed to receive and send signals through the RS485 cable and control the spindle speed. Figure 6.10 shows the front panel of the developed LabVIEW program. The spindle speed is varied from rpm to rpm, which is a marginal increase. The upper bound of the spindle speed is also fixed at rpm considering the safety of the spindle. On-line force measurement X- Direction Resultant Y-Direction Figure 6.10 Front panel of the LabVIEW program for ACC

22 Testing of ACC In order to estimate the capability of the ACC system, experiment along with process capability studies are carried out. Two sets of experiments are conducted. In the first case the ACC is not used, in the second case the ACC is used. Table 6.4 gives the details of the experimental conditions. Table 6.4 Experimental conditions Sl.No Specification 1 Cutting Speed, V c rpm 2 Feed rate, V f 4000 mm/min 3 Depth of cut, b a 0.15 mm/pass 4 Work piece Aluminium - LM6 5 Tool 6mm, solid coated carbide end mill 6 Surface finish (Ra) 1 m, Mean 7 Tolerance on surface finish (HSM) 0.2 to 1.8 m 8 Sample size 60, during gradual wear period of tool The experimental observations are shown in Table 6.5. Both the experimental results are shown in the same table for ease of comparison.

23 120 Table 6.5 Experimental observations ACC system S.No Part No. Cutting force Measured surface Tool Machining (N) finish (µm) Wear Time (s) Without With Without With ( m) ACC ACC ACC ACC

24 121 S.No Part No. Cutting force Measured surface Tool Machining (N) finish (µm) Wear Time (s) ( m) Without With Without With ACC ACC ACC ACC Process capability study The interpretation of capability indices such as C p and C pk rests on a base of several assumptions. The assumptions discussed in Section holds good for this application.

25 122 To find out whether the adaptive control system has resulted in a better process, a process capability study for the processes is devised. The process capability ratio (Cp) is defined in equation (2) Experimental readings from table 6.5 are taken up for the process capability studies. Figure 6.11 (a) and (b) shows the control charts without and with tool wear compensation respectively. (a) (b) Figure 6.11 (a) Control chart (Without ACC) (b) Control chart (With ACC)

26 123 The non uniform trend in the control chart (a) is due to tool wear however from control chart (b) it is clear that each time the surface finish value started drifting on the higher side the increase in cutting speed has significantly reduced the surface finish (Ra) value. From the above control chart, it is observed that the component dimensions are in statistical control. This satisfies the first assumption of process capability analysis i.e., the process should be in statistical control. (a) (b) Figure 6.12 (a) Control chart (With ACC) (Without ACC) (b) Control chart

27 124 From Figure 6.12(a) and (b), it is observed that the components machined without and with ACC system follows normal distribution. This satisfies the second assumption made in process capability analysis i.e., the process should follow normal distribution Comparison of results Table 6.6 shows the comparison between the two processes. Table 6.6 Comparison between the two processes Name of the statistic Three sigma limits Specification limits Without ACC system LCL=0.296 m UCL=1.847 m LSL=0.2 m USL=1.8 m With ACC system LCL=0.369 m UCL=1.604 m LSL=0.2 m USL=1.8 m Process Mean (µ) Standard Deviation (σ) Variance Process Capability Ratio (Cp) Process Capability Index (Cpk) The mean of the process with ACC system has reduced marginally, when compared with the mean of the process without ACC system. The standard deviation and the variance of the process with ACC system are also minimum when compared with

28 125 standard deviation and variance of the process without ACC system, due to less scatter. Also, the process capability indices and process capability ratio have improved with ACC. 6.5 CONCLUDING REMARKS In this chapter, two process-monitoring systems for HSM are constructed. One is a self-adjusting on-line tool wear compensation system and the other is an ACC for improved surface finish. A new method of tool wear compensation using Aerms and surface finish improvement using cutting speed variation have been suggested and successfully completed. This system is found to be more reliable and would result in better dimensional accuracy and surface finish in HSM components. Both the above systems would improve the overall performance of the HSM set-up. Apart from development of theoretical methods, feasibility of the concept has been demonstrated using laboratory experiments.

Computer Numeric Control

Computer Numeric Control TA202A 2017-18(2 nd ) Semester Prof. J. Ramkumar Department of Mechanical Engineering IIT Kanpur Computer Numeric Control A system in which actions are controlled by the direct