Virtual instrumentation in data acquisition and analysis of tool wear monitoring.

Anghel, Alina ; Sarbu, Ionel ; Scurtu, Dan 等

1. INTRODUCTION

Controlling and estimating machining tool wear is one of the aspects of monitoring machining tool state and conditions. This domain is grounds for research to increase the autonomy, productivity and quality of the manufacturing and reduce the economic loss caused by unexpected failures. Among the elements that make up the machine-cutting and device-work piece tools system, the machine cutting tool has the lowest reliability and can cause major defects.

Among the cutting tool wear monitoring methods, vibration analysis plays an important part and the main reason to consider vibration signals for wear monitoring lies in their capability to respond to excitations occurring at sources that are practically inaccessible, like the tool-work piece interface. The ease of incorporating sensors into the machining tool structure, its low price and the better performances comparing with other methods, are among the other reasons. This method is viable in spite of the inherent difficulties.

The methods that process vibration signals to evaluate the tool state are based on the comparison of the average amplitude signal with a reference signal.

The analysis of the signal picked up by the vibration sensor is made in the time domain, as well as in the frequency and amplitude domains to investigate which of these methods is most efficient for the desired monitoring. Estimating the wear state of the turning tool is made by correlating the information provided by the analysis functions.

By designing an entire package of virtual instruments dedicated to the monitoring of the wear of the turning tool, the authors provide a modern and efficient method that can easily replace the traditional monitoring equipment.

This paper's goal is to carry out experimental research using virtual instrumentation on diagnosing tool wear state for turning machining operation by vibration analysis.

The goal of these experiments is to determine a relationship between the tool's condition and the characteristics of the signal picked up by the vibration sensors.

2. CHOOSING THE PROCESSING SCHEME

From the analysis of the tool wear monitoring equipment by vibration analysis, it is made clear that the signal analysis is required in the amplitude domain as well as in the time and frequency domains. There are a variety of methods used, depending on the school and researcher that used them, but many times these methods aren't systematic and cannot be reproduced. The conclusions drawn are often contradictory.

The turning operation was chosen since it is the most widespread cutting operation. The installation of the vibration sensor can be made in the proximity of the tool-work piece contact, so that the detected signal is not distorted.

Monitoring systems that use vibrations as tool-wear information carriers operate using the following general scheme: the complex signal generated in the machining process is captured by a sensor, transformed into an electrical signal and sent to a preprocessing block (containing hardware and software elements) that increases the signal/noise ratio and performs data reduction. The next block extracts the features and, after an analog-digital conversion, carries out signal analysis in the time, amplitude and frequency domains (Dongfeng & Axinte, 2006). Through comparison between the analyzed signal and the reference signal, we can obtain information on tool-wear and identify the trends in its evolution.

In this article, the general scheme for monitoring the wear process is applied using a virtual instrument for acquiring and saving to files the signals received from the sensor. Data acquisition and transducers signal processing were realized using a NI-DAQ board and LabVIEW software (Wu & Liu, 2008).

The file data is processed by a main modulated program, containing four levels: time domain analysis, frequency domain analysis, amplitude domain analysis and dynamic analysis. After the file data is loaded, the analysis domain can be chosen (Fig.1).

In time domain analysis, the cross-correlation function is used. This function describes joint properties of the reference signal and the test signal. By normalizing this function, its calculation is obtained for the two sequences acquired in different manufacturing conditions and envelope detection.

In our program, the analysis in the frequency domain is based on the estimation of the spectra for two signals, the reference and the test signals, with dedicated virtual instruments. Another virtual instrument was designed to display two graphs: the first for the differences between spectra and the second for comparison with the warning and alarm levels. The virtual instrument for frequency domain analysis allows direct observation of the differences between the power spectra of the reference vector and the test vector. Spectra are displayed simultaneously and with contrasting colors. The calculation and the display of the spectra differences allow reading the difference of the spectra amplitude on the vertical axis.

[FIGURE 1 OMITTED]

The two vectors' spectra and the warning and alarm levels are displayed together on the virtual instrument's front panel and eventual equalities or overloads are made visible.

One of the necessary conditions for using these virtual instruments with good results is the correct setting of the warning and alarming levels and the widening number. The high sensitivity of the instruments is such that any warning or alarm situation will be signaled. The risk of false alarms is high if these levels are chosen incorrectly. However, the high sensitivity allows the advantage of being able to observe the tool wear from its very first appearance in the power spectrum. Based on his experience, the operator chooses the multiplying values for the reference spectra to obtain the warning and alarm levels. If these levels are too high then all the negative effects of tool wear will occur and the higher level of maximum wear is observed too late. If the predefined levels are too low, there is the risk of false alarms (Bajic & Lela, 2008).

The amplitude domain analysis displays the evolution of the peak-factor and the trend index, where the peak-factor is the ratio between the maximum and the effective values of the vectors and the trend index is the product of the mean intensity and the mean frequency of occurrence.

The analyzer is a modified complex virtual instrument from the LabView Library. Its adaptation for processing the vectors from the experiments plan enabled the access to different functions and the verification of some of the concluding remarks.

We consider that the possibility of accessing these functions is useful for using the program in other applications.

3. RESULTS AND PERSPECTIVES

Theoretically, it can be considered that the complex signal produced during the machining operation can be described by a function, depending on multiple variables: cutting conditions, dimensions, form and homogeneity of the work piece, tool geometry and so on. Practically, solving this function is very difficult, from mathematical reasons and requires a very large quantity of experimental data. This function is replaced by other functions that partially characterize the process (Bradley & Wong, 2006). For the theme of this paper, five independent variables were considered: revolution, n [rev/min]; feeds, s [mm/rev]; depth, t [mm]; hardness, HB; wear along the clearance face, VB [mm]. For each variable, five values are set up: medium, minimum, maximum and two intermediate values (Table 1). In order to reduce the number of experiments, an experimental project was used with factorial programming, second order, with central rotation. This experimental program reduces the number of experiments and obtains good results because it doesn't lose experiment significance (Dasic & Natsis, 2008). The trend index values seem well correlated with the tool-wear. Increasing these values indicates an active wear mechanism. The practical implications of this remark are the introducing of the trend index in monitoring schemes for other tool wear. For tool wear monitoring, the peaks that trend to zero of the Hilbert transform for the normalized cross-correlation function can be used. The presences of certain tight areas where the mentioned peaks are more wear sensitive, increase.

The modification of a complex virtual instrument from the LabVIEW Library and its adaptation to process the vectors from the experiments plan enabled access to different functions and the verification of some of the concluding remarks.

4. CONCLUSIONS

This paper presents a method of monitoring tool wear using virtual instrumentation that estimates the wear, constructed by wiring together objects that send and receive data, perform specific functions and control the execution flow.

For the research that led to this paper, the authors have leveraged theoretical and experimental methods for tool wear monitoring, in order to apply them to the turning operation

The factorial centered experimental program is also briefly described in this paper, in order to highlight the influence of the tool wear, cutting parameters and hardness of the piece work on the signal that is picked up by the vibration sensor during the turning operation.

A package of virtual instruments operating in the amplitude, time and frequency domains has been conceived to process the vectors containing the experimental data.

The use of this program allows complex processing by selecting different functions and comparative graphical display of the variables. The modular and flexible structure allows usage in other programs and makes it useful for digital signal processing.

The virtual instrumentation that was designed around this topic was utilized in experimental research. The results obtained recommend introducing virtual instrumentation for vibration analysis monitoring and diagnosing.

Other tool state monitoring systems for cutting operations can be based on the research method presented in this paper.

Moreover, future research could associate a greater number of parameters of different nature to establish a technical diagnosis of the machining tool and the machining process.

5. REFERENCES

Bajic, D. & Lela, B. (2008). Examination and Modeling of the Influence of Cutting Parameters in Longitudinal Turning. Journal of Mechanical Engineering, Vol. 54, No. 5/2008, pp 322-333, ISSN 0398-8673.

Bradley, C.& Wong, Y.S. (2006). Surface Texture Indicators of Tool Wear. The International Journal of Advanced Manufacturing Technology, Vol. 17, No. 6/2006, pp. 435-443, ISSN 0268-3768.

Dasic, P. & Natsis, A. (2008). Models of Rehabilitee for Cutting Tools. Examples in Manufacturing Engineering. Journal of Mechanical Engineering, Vol. 54, No. 2/2008, pp 122-130, ISSN 0318-6572.

Dongfeng, S.& Axinte, D.A. (2006). Online Machining Process Monitoring. Instrumentation and Measurement Technology, Vol. 23, No. 8/2006, pp 281-286, ISSN 1091-5281.

Wu, C.Y. & Liu, X.L. (2008). A Study of Cutting State Monitoring System Based on Virtual instrument. Applied Mechanics and Materials, Vol. 10, No 3/2008, pp. 568-572, ISSN 165-6424.

Tab. 1. Independent variables for tool wear monitoring

n [rev/min] s [mm/rev] t [mm] HB VB [mm]

260 0.025 1 165 0
425 0.05 1.25 180 0.15
660 0.1 1.6 200 0.3
850 0.12 2 210 0.5
1040 0.25 2.5 230 0.7