Potential of Radar Based Measuring Systems in Hot Forging

Abstract

Hot forging processes and process chains provide components, which meet highest demands due to mechanical properties. A wide range of applications in different industries like oil and gas, automotive and aviation are based on such processes. As those processes are very productive unfavorable process conditions and unnecessary reject or rework are a strong issues and still not solved for industrial scale production. One key challenge is to generate geometrical information of the workpiece. The geometrical characterization of hot forged parts is usually carried out in cold condition resulting in an extensive time delay to processing. Information about the part geometry during processing and right after processing – in hot condition -- is not accessible. An approach to address this are radar based measuring systems. Those systems have specific characteristics like sufficient accuracy and robustness which qualifies them for measuring tasks under rough conditions typically occurring during forging processes. In the paper, different use cases for the application of radar-based measuring systems are identified and analyses based on laboratory scale experiments.

1 Introduction

This paper presents a systematic analysis of the potential using radar based sensors respectively systems to acquire geometric work piece data in hot forging processes. These processes should be more sustainable designed in the future. The focus criterion there is resource efficiency. Valid data is the basic requirement to optimize processes. Integrated data based optimization has the potential to: a) increase in material efficiency - detection of trends and quality defects; b) increase in efficiency and objectivity in heat measurement c) tighten tolerance of process windows and dimensions d) reduce of raw material mass and temperature e) increase transparency through data collection.

In hot forging, components with a high level of geometric complexity and highest demands on mechanical properties are produced. Forged parts are usually measured after processing in room temperature. Hence, defect parts are identified after complete manufacturing only. Up to recognizing an error in the process, for example systematic ones (die shift, tool wear, etc.), large quantities of rejects are produced and unnecessary additional costs caused. A high potential for industrial relevant component recognition and error detection is the application of radar-based sensor systems. The sensors are characterized by thermal robustness and are able to record the geometry - over time - even in the high-temperature range. These data can be enriched e.g. with temperature information of the forgings.

Various solutions are available for detecting products in manufacturing processes. Depending on the application and the required accuracy, inductive proximity sensors, mechanical tactile solutions, laser-based or optical systems are used. In the forging industry, these solutions usually could not be used effectively due to the requirements and boundary conditions such as cycle times, thermal radiation and possible contamination of the room air with dust and vapors. Other systems focus on measuring the component after forming. Therefore, a hot measuring cell demonstrator based on laser technology is suitable. In principle, the use of laser sensor systems is established for the geometric characterization of warm components. There is a need for encapsulation of the sensors and shielding of the component environment from external influences. Contamination of the optics by dust e.g. However, dust and water vapor, for example, are also unfavorable for the laser-based measuring systems. In addition, the costs of these systems are relatively high.

Radar based technology has been significantly simplified. As a result, radar technology could expand into non-military areas. Radar sensors are increasingly being used for industrial purposes. Distance sensors based on the frequency-modeled continuous wave method (FMCW)are used for production monitoring and quality control [1] Recent developments show that radar-based sensor systems are suitable for thickness and width measurement when used in rolling mills. Therefore integrated, housed sensors like shown in Fig. 1 are developed. This sensor is based on a SiGe radar chip (SiGe = silicon-germanium). The core component of the radar sensor is the circuit board. A shielding body and a dielectric lens are attached to this circuit carrier. The Data interface is Ethernet based. The first pilot plants are operating with reliable measurement results. Other trends are MIMO radar systems (Multiple-Input Multiple-Output) for imaging radar applications as well as imaging radar with rotating sensors, e.g. SAMMI 2.0/3.0 [2].

2 Procedure and Methods

As the goal of the performed work is to identify and verify the potential of radar based measuring systems the following steps are prepared and presented: 1. Identification of relevant use cases in industrial scale (for forgings and preforms); 2. Derivation of scenarios which represent the use cases; 3. Design and setting up of measurement setups; 4. Testing and analysis. A systematical screening of possible applications carried out at different representative process chains is the basis for the application scenarios. As representative processes respectively process chains are the forging processes of gears, of driveshaft flanges, of rings, and of free forged components. The measuring could be performed before, during and after processing. In a next step the identified use cases are clustered and due to similarity of the measuring task summarized into four different scenarios. Table 1 shows the identified scenarios.

Table 1. Use Cases and scenarios

Scenario 1 >>Position<< is designed to investigate and detect a position of a hot semifinished on the tool and the tool position as well. Those use cases address the verification of specimen positioning especially in hand-guided processes, setup processes and tool monitoring. Scenario 2 >>Distance<< relates to measuring of thickness of distinct spots respectively areas, the aspect of spigot length and the aspect of surface and opening detection. Scenario 3 >>Lines<< is based on movement of the sensor relatively to the specimen. It addresses the aspect of radii, the web contour and the diameter detection of components. Scenario 4 >>Surface<< is designed to detect underfil, wear and tool offset by reconstructing a full body of the forged component. For the laboratory analysis scenario 1 to 3 are selected as scenario 4 is at the time not adequately realizable. For the experimental analysis sensors which are successfully applied in hot rolling applications (hot sheet rolling) have been selected. The selected sensor are MS 122-1 FMCW radar from Mecorad like shown in Fig. 1 (left).

The experimental setup used for the investigation of scenario 1 and 2 is depicted in Fig. 1. The main components are two sensors, an NC-milling machine (used for positioning) and an oven. The setup is designed to move the sample in x and y direction. Positioning is determined by the machine control. The calibration of the setup is performed with a distance of app. 1000 mm in x-direction and 2000 mm in y-direction. The change of position is realized in three different increments, in separate series of measurement. Used increments are: 1 mm, 0.2 mm and 0.5 mm.

Figure 1, right shows the two used specimen and the placement of the specimen on the machine table in hot condition. For thermal decoupling an insulating panel is used for initial positioning mounting fixtures are implemented. In hot condition the specimen are heated up for 20 min at 1150 ℃ oven temperature. The surface temperature of the specimen is measured using a pyrometer. A temperature compensation is performed by cooling curves which are generated with the measuring setup. Therefore the specimen are heated up and measured during the cooling on air.

Fig. 1.

Sensor, experimental setup of scenario 1 and 2; hot specimen on table, specimen

As the basic setup shown in Fig. 1 is used for the scenarios >>Position<< and >>Distance<< the compensation curves are used for the evaluation of both scenarios. The >>Distance<<scenario is carried out using exclusively the x-direction sensor. The distance between the sensor and the specimen is app. 1000 mm.

Fig. 2.

Specimen and placement in hot and cold condition

The third investigated scenario is the >>Lines<< scenario. The experimental setup is designed and set up to investigate a rotating specimen linewise. Figure 5 shows the laboratory setup. The key components are rotation axis and a radar-sensor in a distance of app. 1000 mm to the specimen. The experiments are performed with the specimen shown in Fig. 5 as well. The sample is rotated around its own axis in 30-degree increments using a rotating device (see Fig. 2). The rotation started with the value 0°, then the test specimen was rotated in the said 30-degree steps up to 360°. The end position of 360° in turn corresponds to the start position of 0°.

3 Results

The following chapters show the results of the experimental measurings of the selected scenarios: >>Position<<; >>Distance<< and >>Line<<.

3.1 Scenario 1 >>Position<<

Figure 6 shows the distance measurement in the Y direction using a sensor as a function of the displacement of the NC axis in the X direction. There was no displacement of the cuboid in the Y direction. The test evaluation of the position measurement of the side surface of the cuboid shows that this shows significantly higher deviations compared to the distance measurement. The initial value with a displacement of 0 mm represents the maximum value with a distance measurement of 1856.43 mm. The smallest value was detected with a displacement of 6 mm in the x-direction. Here the value of the distance is 1855.14 mm. Within the position measurement, the deviations are in a range of 1.284 mm. This is significantly higher than with the distance measurement. This large systematic measurement error can be explained by the fact that the measurement spot of the radar is too large and thus echo signals were sometimes also recorded outside the measurement body in the event of a shift.

Analogous to the previous measurement, the position measurement of the cylinder lateral surface is shown in Fig. 7. It can be seen that the distance measurement value increases as the NC axis displacement increases due to the curved surface. It should also be mentioned here that a previous calibration took place before the measurement, so that the sensor detected the peak of the diameter contour with a displacement of 0 mm. It can be seen that the deviation increases as the displacement increases. One reason for this strong deviation could again be that the measuring spot was too large. This makes it possible for distance values to be recorded that are located next to the actual ideal measuring point. In addition, it should be noted that the sensor is only able to record a plausible echo signal from the lateral surface of the cylinder up to a displacement of 7 mm.

Fig. 3.

Scenario >>Position<< Evaluation square specimen (left); round specimen (right)

From the test evaluation for the position measurement, it is deduced that the measurement of the straight as well as the curved surface is subject to a significantly higher systematic error. On one hand, this could be due to the measurement of a smaller area in the cuboid measurement, so that the measurement spot was also outside the measured area. In addition, the measurement of the curved surface shows that no echo signal is detected even with a slight inclination of the sensor axis to the test object.

3.2 Scenario 2 >>Distance<<

The results of the series of measurements to determine the measurement error of the radar sensor on a flat surface when cold can be seen in Fig. 4. The individual measurements of the respective increments of the position change of 1 mm are shown here. The blue line stands for the nominal value of the measurements, which corresponds to the displacement of the NC axis in the X direction. The orange dots mark the measured actual value of the displacement of the radar sensor. Here, the actual value refers to the difference between the measured distance and the measured initial value in the starting position of the tool slide. The individual measured actual values shown are again mean values, which were calculated from a set of n = 500. This means that distance measurements were carried out by the radar sensor 500 in a short time for each displacement position of the tool carriage. The results showed the largest absolute deviations in the positive and negative direction of 0.079 mm and −0.105 mm with a step size of 1 mm.

Fig. 4.

Scenario >>Distance<< - measurement square specimen cold, measurement cylindrical specimen cold

To provide the results of the cylinder lateral surface, the results are transferred to a coordinate system analogous to the other distance measurements (Fig. 3). The blue line marks the target value of the measurement and the orange dots indicate the mean value of the displacement measurement of the radar sensor in the respective position. The largest deviation between the target and actual value in the positive range is at a value of 0.149 mm. The largest deviation in the negative range has the value of −0.218 mm.

As addressed in Sect. 3, when determining the displacement, the shrinkage of the cuboid and cylinder in the warm state must be taken into account. For this a shrinkage functions are derived. Based on the determined shrinkage functions and the determined temperature of the specimen, the respective influence of the shrinkage on the displacement can now be determined. Adding the displacement of the axis results in the total displacement, which corresponds to the nominal dimension. This target value is used in the same way as the measurement in the cold state in order to detect the deviations in the sensor measurement. Figure 5 (left) shows the individual results for measuring the side surface of the cuboid in the warm state. As before, the axis shift identity is labeled in blue and the readings in orange. In addition, the grey line characterizes the overall shift. With regard to the absolute deviations, it can be seen that these differ from the respective measurements and in the series of measurements with axis displacement in 1 mm increments, the target and actual values are a maximum of 0.196 mm apart.

Fig. 5.

Scenario >>Distance<< - measurement hot square (left), hot cylindrical (right)

In addition, the investigations into the measurement of the cylinder lateral surface in the warm state were evaluated using the same approach as described for the cuboid. The results are compared in Fig. 5. A similar pattern emerges here. The deviation is the smallest in increments of 1 mm, the maximum deviation here is −0.74 mm. In general, it can be stated that the deviations on the cylinder are significantly larger. A plausible reason for the higher values could be a larger error in the cylinder’s contraction function. This evaluation shows that, under the given conditions, the FMCW radar sensor used can measure the straight surface with an accuracy of approx. ±0.1 mm and the curved surface with an accuracy of approx. ±0.2 mm when cold. This applies to all increments. For the measurement of the cuboid and the cylinder in the warm state, the accuracy of the measurement of the straight surface is approx. ±0.2 mm and the curved surface is approx. ±0.5 mm. Based on these results, it can be deduced that the FMCW radar sensor can detect distances to straight surfaces with higher accuracy compared to curved surfaces. This applies to the cold and warm state of the test specimen.

3.3 Evaluation Line Measurement

A preliminary test with square specimen was carried out to measure the lines of the side surface. A preliminary test was carried out to measure the lines of the side surface. For this purpose, the cuboid was positioned on the rotation device and then rotated around its axis while constantly recording the distance. From this preliminary test it can be deduced that the distance measurement can only measure the distances of a straight surface for small angles of rotation. As can be seen from Fig. 6, measuring a sloping surface facing the sensor, there are strong fluctuations in the measured values recorded by the radar sensor. In the positions 0°, 90°, 180°, 270° and 360°, the surface is perpendicular to the sensor. This shows that the individual fluctuations decrease significantly.

Fig. 6.

Scenario >>Line<< - measurement square specimen

Based on the findings that could be gained from the preliminary test, it was now determined how far the side surface of the cuboid can be twisted. The increment was one degree. The rotation takes place first in the mathematically positive and then in the negative direction of rotation. The steps were carried out until the sensor could no longer record the side faces of the cuboid with sufficient accuracy.

Fig. 7.

Scenario >>Line<< - measurement of a square surface, varying angle of inclination

Figure 7 shows the results of the line measurement in the cold and warm state in a rotation angle range of −7° to 7°. There is a clear tendency here that the measurements become more and more imprecise with increasing tilting. The absolute values tend to deviate strongly from the actual value, which can be determined at 0 °. With increasing rotation angle, after evaluating the data it can also be stated that with increasing tilting (related to the radar device), the standard deviations of the measurements increase significantly in relation to the mean value of the respective individual measurements. It can be derived that the measurement accuracy of the radar sensor with regard to the systematic deviation and random deviation increases significantly when a straight surface, such as the side surface of the test body used, is tilted.

In the case of the cylinder surface, the cylinder was rotated around its own axis in 30-degree increments both in the cold and in the warm state using a rotary device. The rotation started with the value 0°, then the test specimen was rotated in the said 30-degree steps up to 360°. The end position of 360° in turn corresponds to the start position of 0°.

Fig. 8.

Scenario >>Line<< - measurement cylindrical specimen

Figure 8 shows the mean values of the distance measurement between the sensor and the component in the cold and warm state in the respective position. The evalua-tion shows that after the eccentricity has been deducted, the diameter values oscillate in a range between 2.20 mm for the cold measurement and 2.14 mm for the warm measurement. Thus, under the present measurement conditions, the line measure-ment can be carried out equally well in the cold as well as in the warm state.

4 Discussion and Conclusion

Achieving clear reflection of the radar signal and a examinable signal is crucial for a proper measuring using radar based systems. Surfaces that are planar to the sensor can be reconstructed well. The parallel displacement to the sensor can be detected with high resolution for both flat and cylindrical surfaces.

The adaptation of the measuring systems for position monitoring and thickness, width, length measurement appears possible with adaptation measures. The measurement of round material is technically less demanding than billet material. When measuring contours, it has been shown that the inclination of flat surfaces to the sensor already leads to major errors from an angle of ±2°. Inclined surfaces lead to low backscatter. Here, in-depth analyzes and changes to the evaluation algorithms of the echo signal become necessary. The detection of the contours of cylindrical geometries could already be realized quite well. The exact alignment of the sensor to the axis of rotation must be taken into account here, since, as the position detection has also shown, a displacement of the central axis by 8% in relation to the cylinder diameter leads to incorrect signals. Possible solutions are seen here in the trajectory (sensor movement) and further investigations into signal processing steps for better results.

In principle, radar-based measurement technology appears to be suitable for forging applications. Applications are foreseeable, in particular for position detection, distance measurement and diameter measurement.