Although the presented results prove the efficiency of radioscopy, this system have certain characteristics which justify to develop and employ further process integrated testing systems. One of this characteristics is that the integration of radioscopy in industrial applications is doubtful because of reasons of radiation protection. This means, that the results from radioscopy should rather be used to fit other systems (acoustic emission analysis or temperature analysis) for industrial applications.
Detection of near-surface casting defects with temperature analysis
Fig 5: Defect detection with temparature analysis
Fig 6: Temparature courses of child mold and casting part(above); Acoustic emission signal of the casting(below)
| |
A simple, on industrial applications orientated system represents the temperature analysis of critical child mold areas. Those critical areas (hot spots) are controled by thermoelements which are integrated inside the child mold (Fig. 4). Monitoring the changings of the temperature gradients over the time enables to detect casting defects which are situated near the surface around the hot spots. Because of volume deficits like internal shrinkages or hot tears (Fig. 5) the heat flow from the casting part to the child mold near the surface will be disturbed and the determined temperature gradient will take high negative values (Fig. 5).
Detection of casting defects by combination of radioscopy and accoustic emission
More difficult it is to analyse and value the acoustic emission sig-nals which are origina-ted from the casting part and the child mold during the casting pro-cess (Fig. 6 below). Only a analysing method which takes both systems (radio-scopy and acoustic emission) into consi-deration will give the possibility to correlate extracted quality cha-racteristics like the ca-sting part density with characteristic values from acoustic emission.
Analysis of radioscopic values
For the objective and quantitative description of dynamic events of the casting part during the filling-, solidification- and cooling phases a automatic detection of casting defects is necessary and was realized in four partial steps (Fig. 7 left).
Fig 7: Correlation of radioscopy and Accoustic emisssion analysis
|
Image recording
Because of reasons of radiation protection the filling of the test child mold which is placed between the microfocus X-ray tube and the image identifyer was realized with a remote controled casting system. In the first partial step the image values generated by the image identifyer were recorded with a conventional CCD-video and sent to the image processing system for digitizing, integration and storage.
Image pre-processing
Due to the spherical form of the input screen of the image identifyer tube and electromagnetical distur-bances the image values are globaly and localy distorted. Therefore they have to be equalized in the second partial step. The exigency of this pre-process step, which is particulary described in /2/, results from the intention to evaluate quantitative the defects which become visible in the radioscopy image. After the image equalizing, operations like substraction and gray-value modification to minimize inhomogeneous illumination will be the following. The result of the image value pre-processing is a equalized and illumination corrected image series
Defect Segmentation
The aim of defect seg-mentation is to delimit the detected defect as precise as possible from the background which surrounds the defect (Fig. 8). The extracted characteristics of a de-tected casting defect depend considerable on the threshold value which is used during the defect segmentation. Therefore a automatic method, which means a objective and reprodu-cible process, is necessary to determine the threshold value. The results of this investigations show that the threshold value can be determined reproducible in the point of intersection of two normal distributed frequency approximations.
The extraction of charac-teristics contains the de-termination of properties and characteristics which describe the casting de-fect. Within this investi-gations three different groups of characteristics were determined which include informations about the geometry, the distribution of gray values, and the contour of the defects (Fig. 9).
FIg 9: Conception for the extraction of characteristics
|
To apply the ex-traction of characteristics not on single images but on series images will give the possibility to repre-sent and analyse the cha-racteristics in depen-dance on time. As it is shown on one geometric characteristic, the casting defect area (Fig. 10), growth processes of dif-ferent casting defects can be analysed specific-ly. The curves in figure 10 show clearly for each type of defect characte-ristic courses.
Fig 10: Area of occuring defects in dependence on time
|
Aditional to the in-vestigations with the radio-scopy system the Institute of Nuclear Engineering (COPPE) made investigations in cooperation with IKPH using a tomography system.
Fig 11: Tomography scans of a internal shrinkage at different depths(A;B;C)
|
The radioscopy gives only information in two-dimensional form which means that it is impossible to get informations about the defect extension into the third dimension. Therefore investigations using tomography were made. Some results of the tomography images are shown in figure 11. The characteristic values given by analysing the tomography results were the defect volume and the placement coordinates of the defect inside the casting part.
Analyse of acoustic emission values
To analyse and evaluate the acoustic emission signals send out from the casting part and the child mold, which means to determine informative characteristic values a conception was realized which contains as well four partial steps (Fig. 7).
Recording of values
The recording of the acoustic emission values is realized with a conventional PC with a 120MHz Pentium processor, 64Mbyte RAM and a integrated eight channel analog/digital converter card. One-channel measurings permit a maximum sample rate of 700kHz which is equal to a recording time about 46s. For low frequency investigations up to 50MHz analysing frequency, conventional piezoelectric acceleration trancducers were used.
Data pre-processing
Due to the high amount of data, which means a maximum data memory of 128Mbyte per measuring in case of basic investigations, the first step of the data pre-processing was to extract the data sections concerning the origination of casting defects. The informations which are necessary to realize the extraction are contained in the synchronously recorded and already processed radioscopy image values. To eliminate occuring high- respectively low-frequence disturbances the extracted signal sections which are relevant for further analysis will be digital high- respectively lowpass filtered (for example low-frequence disturbances are caused by dispersion effects of the power supply, high frequence disturbances are caused by noise pulses from the thyristor control).
Calculation of characteristic values
In the third partial step, a special analysing software was employed to determine characteristic values from the extracted signal sequences. Due to this characteristic values which describe the form respectively the behaviour of the extracted signal sequences can be calculated in time- and frequency-domaine. For example changings of signal intensities can be described by statistical time based characteristic values as the variance or standard deviation. The curtosis factor characterizes the peak containment inside the signal course. In the frequency domain the autospectral power density describes each component of the time signal in frequence and amplitude. Due to this compression of information at the same time a reduction of data can be realized about factors in the range of 103 up to 104.
Extraction and reduction of characteristic values
The calculation of characteristic values causes a high amount of values which contain redudant informations. Due to this the forth partial step will be to reduce this amount of values using extraction methods. This can be realized with statistical methods like cross correlation analysis.
Correlation of radioscopy and acoustic emission
The last part, following the method to analyse radioscopy- and acoustic emission values, will be to correlate the characteristic values of the radioscopic detection of casting defects with extracted characteristic values of the acoustic emission analysis.
The correlation between the time based characteristic values of acoustic emission analysis and the defect characterizing radioscopy values did not come to very satisfactory results referring the low-frequency measurements. The reason can be found in the fact that different acoustic emission signals initiated from different events were superimposed. Events like the origination of casting defects can occure at the same time with events like e.g. thermal caused relative motions between the casting part and the child mold.
Fig 12: Time based amplitude spectra of a defect free casting part
|
For the distinction of these superimposed acoustic emission sig-nals it is advantageous to make a time dis-integrated investigation of the signals analysing them in the frequency domain. Reference in-vestigations on defect free casting parts showed that the time based amplitude spec-tras of the acoustic emission signal consists only of low-frequency parts (Fig. 12). The reasons for those low-frequent emission events are flow processes du-ring the child mold filling or friction processes bet-ween the casting part and the child mold.
Investigations
Fig 13: Time based amplitude spectra of a casting part with a hot tear in correlation with radioscopy images
|
where, using radiosco-py, hot tears could be detected show that the time based amplitude spectras contains beside the known low-frecuency parts also higher frequency parts (Fig. 13). That means, the results from radioscopy could be correlated exactly with the results from acoustic emission analysis.
Fig 1: 
Fig 2: 
Fig 3:
Fig 4: 
Fig 5: 
Fig 6: 
Fig 8:
Fig 11:
Fig 13: