Thermal Compensation Algorithm For Machine Tool

Transcript

Proceedings of IMECE”05 2005 ASME INTERNATIONAL MECHANICAL ENGINEERING CONGRESS AND EXPOSITION Orlando, FL, USA November 5-11, 2005 IMECE2005-79768 THERMAL COMPENSATION ALGORITHM FOR MACHINE TOOL. E. Kushnir, Hardinge Inc., Elmira, NY ABSTRACT Thermal effects can contributed more than 50% of the overall error in a machine tool and as precision requirements continue to grow the necessity of reducing this error source is essential. Today’s CNC allow to perform indirect compensation of these errors based on temperature measured in different points of machine tool structure. The efficiency of thermal compensation may be measured as a ratio between capability of machining process before and after applying thermal compensation. The proposed methodology of thermal compensation algorithm development is based at testing and FEA. The efficiency of proposed approach is illustrated by compensation algorithm, developed for different type of lathe. The application of thermal compensation allowed to increase capability of machining process more the two times. The accumulated experience in thermal compensation development provides ways for further improvement of lathe machining process capabilities. Keywords: machine tool, thermal deformation, accuracy, thermal compensation. INTRODUCTION Thermal effects can contributed more than 50% of the overall error in a machine tool and as precision requirements continue to grow the necessity of reducing this error source is essential. Optimizing the design of the machine tool is the most effective way to reduce thermal errors. Design measures alone are insufficient, however, to completely avoid thermal displacement. Compensation techniques are essential for increasing the working accuracy and the power of today’s CNC enables these compensation methods. There are two methods for thermal error compensation: - Direct compensation: by measuring the error at every part and compensating accordingly; - Indirect compensation: in this case the thermalelastic deformations are computed for various temperatures, measured in different points of the machine structure. The calculated deflections are then compensated by means of the CNC. It was discussed in publications that “ the complete temperature distribution in machine must be known for a precise indirect compensation”, in practical means this is not possible. The real question is what are the limits of indirect compensation if the temperature is measured only at a few points of a machine, and what is a tolerable error for this type of compensation. 1 Copyright © 2005 by ASME In modern manufacturing the tolerance requirements for a finished part are usually transferred to a requirement for a manufacturing process. These manufacturing requirements are expressed in statistically terms as capability of manufacturing (machining) process. It means that requirements to thermal compensation system have to be evaluated statistically. The efficiency of thermal compensation may be measured in this approach as a ratio between capability of machining process before and after applying the compensation. A machining process in terms of accuracy may be measured in deviation of machined parts size versus time (part number, or machining cycles). The term continuous machining accuracy (CMA) is often used to describe parts diameter variation. Fig.1. Lathe structure. Other source of coolant temperature change is heat developed in coolant pump and pumps motor, and by resistance in the coolant feed lines. The coolant is in direct contact with machine tool structure at different locations. The coolant tank, which is typically located under the base and on the sides, has a large area of contact with structure. The coolant in the tank effects the base directly by heat exchange and by vapors. The change in coolant temperature is one of the main factors effecting thermal deformation of the machine. The majority of mechanical losses are presented by losses in spindle bearings. The heat developed in the bearings will migrate through the headstock. The headstock growth will effect the position of the spindle relative to the top of the base. The heat developed in the collet closer (hydraulic losses) has similar effect. In the case of wrap-around motors the heat developed by electrical losses is added to the heat developed by bearing, and can result in further displacement of the headstock. This effect may be reduced by a chillier, used to cooling up coolant pumped through a wrap-around motor. In the case of spindle driven by belted motor, the majority of the heat developed by the motor will go in the base. Other components effected by mechanical losses are ball screws. The ball screw grows proportional to the increase in temperature and this growth may be compensated by a linear scale, screw pretension, GENERAL OBSERVATION OF A MACHINE TOOL STRUCTURE MOTION UNDER THERMAL LOAD. In precision machine tools heat comes from different sources. 1. Heat released from cutting process. This heat is distributed among the work piece, the chips, and the coolant. The majority of heat (60-80%) is carried away by chips, and eventually will be transferred to the coolant. 2. Heat developed due mechanical, electrical and hydraulic losses. The mechanical losses are represented by motion (spindle bearing) losses. The electrical losses are represented by heat developed in motors and drives. The hydraulic losses are represented by fluid dynamic friction energy losses in collet closer, hydraulic and coolant pump. 3.Thermal effect of change in the ambient environment. By analyzing these heat sources we can evaluate the possibility to control or stabilizing their effect on a machine tool structure motion. For example, in the case of lathe (see Fig.1), heat released from cutting process primary will be transferred to the coolant and increase its temperature. 2 Copyright © 2005 by ASME information for theoretical compensation problem. and through balls crew coolant or by separate thermal compensation system. As it follows from presented observation there are two main structural items in lathe that are most effected by heat from the machine internal sources: headstock and base. External sources of heat may be divided in two categories: short and long-time effect heat sources. General machine tool structure has time constant in the range of 1.5 – 3 hours. The short time effect sources are sources that have time constant less than time constant of a machine tool structure. Long-time changes in the environment, as for example, change in a shop temperature during a day, have time constant longer than one of machine tool structure. The short time effect sources included currents of warm and cold air, short time variation in shop temperature because of opened door, etc. This type of temperature variation in the environment will develop transition processes in the structure components with different delay time between change in environment and thermal displacement of machine components. It is very difficult and expensive to follow this variation in environment. The best approach to this problem is to have a relative stable shop environment in which all the changes have time constant higher than 1.5 – 3 hours. That means a lathe with thermal compensation system has to work in the environment with long time effect external sources of heat. In this case the change in the environment will effect temperature of the coolant and main components of the machine structure. These temperatures may be used by the thermal compensation system as information points. study of thermal 0 80 -0.0002 76 -0.0004 72 -0.0006 68 Tim e, m in 0 50 Headstock Base Front DIA var, in 100 150 200 250 Front spindle bearing Base rear 300 DIA variation, in Temperature, F Test data. CMA duty cycle. 84 -0.0008 350 Coolant Base Top Fig.2. Temperature and part diameter variation data from lathe CMA test. The FE model of a lathe structure was developed and used in this study. It is very difficult accurately to define thermal load and boundary condition for the FEA thermal model. Therefore, the FE analysis was used only for quality evaluation of structural behavior with emphasis to define points for temperature measurements at the next step of development of thermal compensation algorithm. For this reason a simplified three-step approach was used to set up the condition for structure thermal displacement evaluation. PRELIMINARY TESTING AND THERMAL ANALYSES OF LATHE STRUCTURE. As it follows from previous observation temperature distribution in base and headstock will have major effect at motion of machine tool structure. For this reason in preliminary continuous machining accuracy (CMA) testing the temperatures of the base and headstock in different points were recorded. The results of one of these tests are presented in Figure 2. This data were used as a basic 3 Copyright © 2005 by ASME Step1. A simplified temperature distribution presented in the Figure 3 was used as thermal load at the structure. The relative displacements between spindle and tool point under different thermal loads (from headstock only, from base only, from both sources combined) was computed. It followed from this simply temperature distribution models: - the temperature distortion in the base causes movement of the headstock; - the assumed non-uniform temperature distribution in the base may decrease relative displacement in the cutting zone, because complex twisting of the base causes the carriage to rotate away from the headstock. Step 2. More complicated temperature distribution was developed to verify the effects observed at step 1. At this step the heat sources were put in headstock and base and temperature distributions in the structure were observed at different moments in time. The moment, when temperature distribution was similar to data obtained in preliminary test, was chosen for relative displacement evaluation. The results of these analyses showed that the heat from coolant causes the base to move away from headstock and decrease part diameter variation. Step 3. A thermal load, similar to one used at step two, was applied as a function of time. The goal was to get temperatures as functions of time with profiles, similar to that obtained in tests in specific points of the machine. The example of temperature distribution in the structure in one particular moment of analyses is presented in Figure 4. The changes in temperature in different points of structure and calculated value of part diameter variation are presented in Figure 5. 0 -0.0002 DIA variation -0.0004 76 Headstock -0.0006 Base front 72 Base rear -0.0008 68 0 50 100 150 base front under riser riser (joint with head) 200 250 300 Time, min base rear DIA variation, in Temperature, F 80 -0.001 350 (middle up) DIA variation Fig. 5. Theoretical temperatures and part diameter variation as a function of thermal load. It follows from Figure 5 that assumed thermal load developed parts diameter variation, similar to one obtained in cutting tests (see Fig.2). The performed FEA analysis showed that under simple assumptions about temperature distribution in lathe structure, deviation in cutting part diameter will be similar to one obtained in testing. These results can be used to choose points in the structure where temperatures have to be measured for adequate control of the relative displacement in the cutting zone. As followed from presented data, these points have to be situated at least in followed areas of the structure: headstock, base below headstock, base near coolant tank, rear of the base. THERMAL COMPENSATION MODEL DEVELOPMENT The goal of non-direct thermal compensation is to correct thermal errors using measurements of the temperature at multiple locations on the machine tool. This goal has to be reached with minimum expense, which means with use of a limited number of relatively expensive temperature measurement channels. The Fanuc CNC offers thermal control modules with 4 channels. Therefore, the preliminary task is to find not more than 4 points on the machine tool structure, which will represent effect of thermal distortion at the displacement between the tool and 4 Copyright © 2005 by ASME - the relative rapid change of the temperature at collet closer and in hydrostatic oil make it impossible to use these points for compensation control; - base rear and base top points have very similar behavior; - headstock, base front, base under headstock, and headstock front bearings behave almost the same way. work piece. For example, consider algorithm development for the lathe with hydrostatic guides and high-speed wrap around spindle. The 4 points were chosen in two step process: 1-preliminary testing with maximum 12 thermal sensors including points, chosen as the results of FE analysis; 2-correlation analysis to choose 4 or less thermal channels. The position of thermal sensors at the first step was defined based at following 4 criterion: - results of FE analysis; - thermal sensors have to be clamped to components that will average relative fast changes in temperature. For example, they have to be clamped to pieces of cast iron or put in large volumes of oil to avoid rapid temperature variations. - thermal sensors have to be clamped close to points where the largest amount of heat is developed and transferred. - thermal sensors have to be distributed in the machine tool structure to represent its major thermal components, which define displacement in cutting zone. The position of preliminary points of temperature measurement were chosen as follow: 1 - base under headstock; 2 - headstock side wall; 3 - headstock near front bearings set; 4 - base front; 5 - base top; 6 - carriage up; 7- carriage low; 8 - collet closer; 9 - base rear; 10 - hydraulic oil tank; 11 - hydrostatic oil tank; 12 - ambient. The results of measurements are presented in Figure 6. In general, the correlation analysis has to be performed to chose independent points for thermal compensation model. Fortunately, in many cases some of the points may be eliminated based at simple evidence and observation of test data. It follows from Figure 6 that: 40 Temperature, C 36 1 Base(under head) 3 Carriage low 5 Hydrostatic oil 7 Base rear 9 Base top 11 Riser 32 2 Hydrolic oil 4 Carriage Up 6 Collet 8 Head 10 Base front 12 Ambient 28 24 Time, min 20 0 43 85 128 170 213 Figure 6. Change in temperatures during CMA test. It means that there are only 5 points to consider for use in thermal compensation model: 1hydraulic oil; 2base under headstock, as representative of headstock, base front, and headstock; 3carriage low; 4carriage up; 5base rear, as representative of base. As it was mention earlier CNC thermal control module has only 4 inputs. The decision what point to drop may be based on correlation analysis of data from these 5 points. The correlation matrix of data from these 5 points is presented in the Table 1. The highest correlation occurs between “Base rear” and “Carriage up” points. Thus, only one of this point, for example “Carriage up” might to be used in the compensation model. Obtained from this analysis 4 points: “Base under headstock”, “Hydraulic oil”, “Carriage up”, and “Carriage low” were used in compensation model developing. A similar approach, used for lathe with linear guides and belted spindle, left only two independent points for thermal compensation model: “Base under headstock’ and “Headstock side wall”. 5 Copyright © 2005 by ASME range 0”- 0.0013” in the range of +/- 0.00015” (see Figure 8). The lathe with hydrostatic guides and belted spindle required 4 points: “Base under headstock”, “Headstock side wall”, “Carriage Up”, and “Carriage low”. CMA of a lathe with hydrostatic guides and wrap around spindle. Temperature variation in the chosen points 9 Base Carriage low 7.5 Tempearture, Table 1 Measurement points Correlation matrix Base Hydrau Carria Carria Base (under lic oil ge low ge Up rear head) Base 1 0.629 0.876 0.923 0.905 (under head) Hydraul 0.629 1 0.914 0.852 0.864 ic oil Carriage 0.876 0.914 1 0.991 0.99 low Carriage 0.923 0.852 0.991 1 0.995 Up Base 0.905 0.864 0.99 0.995 1 rear 6 4.5 3 1.5 Time, min 0 0 33 65 98 131 163 196 229 261 294 327 Fig. 7. Temperature change during CMA test. CMA of a lathe with hydrostatic guides and wrap around spindle 1.60E-03 1.40E-03 DIA variation, 1.20E-03 Cutting data Compensation model Accuracy predicted by model 1.00E-03 8.00E-04 6.00E-04 4.00E-04 2.00E-04 0.00E+00 -2.00E-04 0 The regression models of different orders may be used as a thermal compensation model. Let’s consider the simple regression linear model with 4 variable temperatures T in chosen points as an input variables and diameter variation (DIA) as output variable: DIA=C1*T1+C2*T2+C3*T3+C4*T4 Hydrolic Oil Carriage Up 50 100 150 200 250 TIME, min 300 350 Fig.8. Part diameter variation. Test data and results of thermal compensation. Data in Figure 8 showed that indirect compensation might be performed based at the correlation between relative displacement in cutting zone and temperatures in chosen points of machine tool structure. (1) In this model C1, C2, C3, and C4 are coefficients of regression defined by least-squares procedure based at test data. Figure 7 presents temperature data obtained during testing of a lathe with hydrostatic guides and wrap around spindle. Parts diameter variation obtained during the same CMA test is presented in Figure 8. The parts were cut every 8-th cycle. Between the cuts machine was performing the same duty cycle, including running coolant. The regression analysis showed that with coefficients C1= 12, C2=-4, C3=8,C4=-6 diameter variation may be brought from THERMAL COMPENSATION PROCESS CAPABILITY. The capability of machining process is defined statistically. It means that the process has to be capable to delivery required accuracy not only through one run but day after day (over long period of time). It means that the parameters of the model, that are determined based at one or more particulars runs, have provide required accuracy during other runs and in different environment conditions. In other words these parameters have to be nonsensitive to changes in environment and cutting conditions. 6 Copyright © 2005 by ASME Follow ASME Standard E-2281-03 “Standard Practice for Process and Measurement Capability Indices” we will measure machining process variability over long period of time by process performance PP or process spread as: PP = 6 σLT , where σLT – overall sample standard deviation. In process performance, the actual performance level of process is estimated rather than its capability when it is in control. That makes this index very suitable for comparison of machining processes before and after applying thermal compensation. The standard CMA test was set up to check the sensitivity of the regression model parameters on a lathe with linear guides and belted spindle motor. The test was set up to be run unattended and repeated 7 days in the row. The relative displacement between workpiece and tool was measured by capacity probe (accuracy +/- 0.000020”). Acquisition system was used to collect information about relative displacement. The obtained CMA records are presented in Figure 9. The performed regression analysis with the data obtained during testing of the machine with linear guides (see data in Figure 10) showed that set of factors in Eq. 1: C1=-10, C2=24, C3=0, and C4 =0, simultaneously brings diameter variation in all 7 test in the range +/- 0.00027” (see Figure 10). DIA variation of CMA without compensation The diameter variation distribution, obtained by this transformation is presented in Figure 11 as the results after thermal compensation. Process performance index PP for this test equal 6*0.00011'' = 0.00066’’. As it follows from the presented data the manufacturing process, obtained by thermal compensation increase machining process performance 1.82 times. Ac c ura c y of CM A Te s ts w ith The rm a Com pe ns a tion 0.000500 0.000400 0.000300 DIA var, -0.000300 -0.000400 1 TEST#1 DIA var, 0 -0.0002 -0.0003 -0.0004 TEST#1 Test#5 Test #2 Test #6 33 41 49 Test#3 Test #7 57 65 73 17 25 Tes t #2 Tes t #6 33 41 Tes t#3 49 57 Tes t#4 65 73 81 Tes t #7 Fig. 10. Predicted part diameter variation for all 7 runs. -0.0001 25 9 Tes t#5 0.0001 17 Cy c le # -0.000500 0.0002 9 0.000000 -0.000100 -0.000200 Data are presented relative to average value for 7 runs 1 24 0.000100 0.0003 -0.0005 -10 C2= 0.000200 0.0005 0.0004 C1= Cycle 81 Histogram of DIA variation Test#4 Frequency Fig. 9. Part diameter variation in 7 CMA tests, represented relative to average in all 7 runs. For an easy comparison with thermal compensation results the data at Figure 9 are presented relative to average value of diameter variation (DIA) for all 7 runs, chosen as a zero diameter variation point. As it follows from Figure 9 diameter variation in all 7 test is in a range +/0.00045”. Process performance index PP for this test equal 6*0.0002=0.0012”. Part diameter distribution for this process is presented in Figure 11. 140 Original Data 120 Results after compensation 100 80 60 40 20 0 -0.0005 -0.0003 -0.0001 0.0001 0.0003 0.0005 DIA variation, in Fig.11. Part diameter distribution before and after applying thermal compensation. Similar tests were performed on a lathe with hydrostatic guides. Tests were performed with 7 Copyright © 2005 by ASME Histogram of DIA variation (Lathe with Hydrostatic guides) cutting of actual parts every 8-th cycle. In Figure 12 are presented results of parts diameter variation obtained during 5 CMA runs on a machine with hydrostatic guides. 12 Original data Frequenc 10 1.6E-03 Results after compensation 8 6 4 1.4E-03 2 DIA variation , in 1.2E-03 0 -0.0011 -0.0008 -0.0005 -0.0001 0.0002 0.0005 0.0008 1.0E-03 8.0E-04 DIA variation, in 6.0E-04 Fig.14. Part diameter distribution before and after applying thermal compensation (lathe with hydrostatic guides). 4.0E-04 Test #2 Test #6 2.0E-04 Test #3 Test #7 Test#4 Time, min 0.0E+00 0 30 60 91 121 151 181 211 241 271 301 Figure 12. Parts diameter variation obtained during CMA runs on a machine with hydrostatic guides. As it follows from comparison of the curves in Figure 9 and Figure 12 these machines behave significantly different. The performed regression analysis with the data obtained during testing of last machine showed that set of factors in Eq. 1 C1=-38, C2=12, C3=36, and C4 =3, simultaneously brings diameter variation in all 5 test in the range +/- 0.0003” (see Figure 13). The original process (without thermal compensation) has process performance index PP for this test equal 6*0.000405” = 0.00243”. That means that actual performance of the process with thermal compensation is 3.25 times better. THERMAL COMPENSATION ALGORITHM SENSITIVITY. Thermal compensation algorithm sensitivity to parameters may be found by comparison of process performance estimated for algorithms based at different runs. For example, in Table 4 are presented results of efficiency of thermal compensation for different sets of parameters used for thermal compensation in the case of test data presented in Figure 10. Every set of parameters was chosen as the best for particular run and than the efficiency of thermal compensation was estimated for all 7 runs. Predicted DIA variation relative to 5 runs average Factors(-38,12,36,3) 4.0E-04 DIA variation, in 3.0E-04 2.0E-04 1.0E-04 0.0E+00 -1.0E-04 -2.0E-04 -3.0E-04 -4.0E-04 0 Test #2 30 60 91 Test#3 121 151 Test #4 181 211 Test #6 241 Time, min 271 301 Test #7 Fig.13. Predicted diameter variation after applying compensation (lathe with hydrostatic guides). The diameter variation distribution, obtained by this transformation is presented in Figure 14. The manufacturing process, obtained by thermal compensation, has process performance index PP for this test equal 6*0.000124 = 0.00075”. 8 Test # C1 C2 1 2 3 4 5 6 7 best for 7 -8 -11 -11 -10 -11 -5 -10 -10 16 20 29 20 22 30 28 24 Table 4 Efficiency of Thermal compensation 1.52 1.66 1.89 1.67 1.74 1.84 1.88 1.82 Copyright © 2005 by ASME As it follows from presented data parameters of algorithm, defined by any of performed runs, represent overall machining process with sufficient accuracy. It means that algorithm is not very sensitive to accuracy of parameters. These results opened the way for practical application of developed approach in machine tool industry. application model allow to perform thermal compensation in different type of lathe. This compensation algorithm has the ability to increase continuos machining accuracy of a lathe from1.5 to 3.25 times. In presented form the algorithm is implemented by one of machine tool builders in its products. CONCLUSION Presented test data shows that proposed methodology of indirect thermal compensation algorithm development and obtained from it REFERENCES 1. “Standard Practice for Process and Measurement Capability Indices”, ASME Standard E-2281-03 9 Copyright © 2005 by ASME