4.1 Construction

The sensing element consists of three layers: a low resistive layer and a high resistive layer separated by a soft material as shown in Fig. 4a. Initially, a five-layer carbon-fiber composite and a two-layer glass-fiber composite were manually fabricated. After the curing of the composite plates, the top epoxy surface of the carbon-fiber composite was removed to make it electrically conductive by a hand-sanding process. After removing the epoxy layer, the top surface can be used as a low resistive layer. For a high resistance layer, a resistive thread was stitched on the glass-fiber composite using a sewing machine. The resistive thread consists of 65 % silk and 35 % stainless steel (AISI-316L grade) material by mass. In the end, both the layers were joined together using a general silicone sealant. The silicone sealant was applied to form the hollow rectangular shape on carbon-fiber composite as shown in Fig. 4a. Then the high resistive layer was placed over it and allowed to cure under the static load at room temperature. Figure 4b explains the manufacturing process used for preparing the sensor prototype.

Figure 4CFRP-based validation sensor: (a) construction, (b) manufacturing process and (c) resistance diagram of sensor at different contact points.

Download

For testing purposes, thin plate-like sensor prototypes were manually prepared in the laboratory. The final goal of the research work is to integrate the sensing concept in the vehicle parts like the front bumper, which is the final product, as shown in Fig. 3.

4.2 Working principle

Figure 5 shows the electrical schematic diagram of the sensor. A constant voltage is applied across the high resistive layer in series with a shunt resistance. The shunt resistance is connected to measure the current flowing through the sensor. The voltage across the resistive layer U_in, the current through the circuit I_in (which can be calculated from U_shunt and R_shunt using Ohm's law) and the measured voltage drop U_m over the known resistance (R_known=1 MΩ) are continuously measured with respect to time.

Figure 5Electrical schematic diagram of sensor: (a) no crash, (b) single-point contact and (c) multi-point contact.

Download

During the normal driving condition, there is no electrical contact between the high resistive layer and the low resistive layer. Therefore, the measured voltage drop U_m is almost zero. If a crash occurs, the resistive layer is deformed, which causes the two layers to come in contact with each other. As soon as the contact occurs, U_m and I_in change depending on the position and the width of the contact (overlap). There are two types of contact which can take place during the crash event: single-point contact (Fig. 5b) and multi-point contact (Fig. 5c).

5.1 Experimental setup

Figure 6 shows the schematic of an experimental setup used for the pendulum tests. It consists of the pendulum structure, the pendulum rod with the additional impacting mass, trigger (light gate), data acquisition system (LTT24, Labortechnik Tasler GmbH) and power supply. The sensor prototype was fixed on the fixed beam of the pendulum structure. The pendulum rod was attached with an additional mass of 11 kg to ensure that the impact energy is high enough to deform the prototype. A wooden block was fixed on the pendulum arm which acts as an impacting surface. The pendulum rod, along with the wooden block and mass, was locked at an angle and was released to impact the sensor. A light gate was used to trigger the data acquisition system, when the wooden block first touches the sensor.

Figure 6Pendulum experiments: (a) test setup and (b) test configurations.

Download

The sensor was tested with impactors (wooden block) of two different widths (10 and 20 cm) at three different impact positions each. The impact speed for all tests was approximately 4.2 m s⁻¹. Figure 6b shows the different tests performed using the pendulum equipment.

5.2 Results and discussion

The raw signals U_in, I_in and U_m are measured with data acquisition system with a sampling rate of 10 kHz, which is triggered using the light gate. These raw signals are converted into real-time values of position P and overlap O using Eqs. (7) and (5) respectively. The position and the overlap signals are illustrated in Fig. 7.

Figure 7Results from pendulum tests: (a) different impact position (with 10 cm impactor) and (b) different overlap (at impact position 37.5 cm).

Download

Table 2Results of pendulum tests.

Download Print Version | Download XLSX

The tests where the sensor impacted at different positions show the capability of the sensor to determine the position of the impact as demonstrated in Fig. 7a. The actual position (corresponding to the midpoint of impactor) and actual overlap values were measured with a measuring scale before the test. In addition to the pendulum motion, the pendulum arm also had minor twisting motion. This caused an initial contact on the edge between the impactor and the sensor, followed by a surface contact (as seen by the signals in Fig. 7). Hence, for the estimate of the position value, an average value of the position signal over a time period of 3 ms was considered. As soon as the contact occurs, the position value rises very steeply (400–1000 m s⁻¹) from zero until the position of the contact point is reached. Then the slope of position signal depends on the contact width or overlap development (which is comparatively low at about 100–150 m s⁻¹). Therefore to select the starting point of average time period, a time instance with considerable change in the slope of the position signal was chosen. For the overlap estimation, the maximum value of the overlap signal was considered. During the rebound phase (when the impactor travels back after the impact), it was observed that the overlap values are negative. The negative values of overlap are realized due to the drop in I_in, which is caused by the increase in resistance of the sensor because of tension loading on the resistive thread. When the high resistive layer is in contact with low resistive layer, the above effect is negligible, because the tensed part of resistance thread is in contact with the CFRP-layer (low resistive layer) of the sensor. As this effect occurs during the rebound phase, this effect is not of importance for a real crash scenario because a safety decision should be taken within the first few milliseconds. The values of actual position and overlap for different tests are displayed in the Table 2. The table also lists the corresponding sensor-estimated values for different tests. The relative error for both position and overlap was calculated using the following equation:

Table 2 summarizes the results from the different pendulum tests performed. It can be seen that the position values can be determined with a relative error below 12 % at 3.9–5.3 ms after T0. Figure 7b displays the results from two tests with different overlap, which demonstrate the potential of the sensor to provide information on the overlap. The average relative error of overlap estimation is about 30 %. The estimation of overlap requires some time for the deformation of the sensor until the width of the contact is equal to the width of the impactor. Hence, the overlap can be estimated at about 9–15 ms after T0 (at 4.2 m s⁻¹). This time will reduce with increased impact speed and therefore faster deformation. In a real crash with higher energy and the sensor installed directly on the bumper, it might be that the sensor is (partially) broken before the complete overlap is measured. Also the development of the overlap will be based on the curvature of the vehicle bumper. But some information is still better than no information, and the available information can be used to validate the overlap development for some samples, as explained in Sect. 6.3.

6.1 Crash test configuration

Figure 8a shows the crash test setup, and Fig. 8b displays the positions of the different sensors installed in the vehicle during the crash test. The vehicle used for crash test was a Volkswagen Golf IV. The crash test was carried out at 64 km h⁻¹ against a rigid wall with 40 % overlap (40 % of vehicle width will come in contact against the barrier). The vehicle was aligned in such a way that the impact occurs on the driver side. Two types of acceleration sensors were used. A triaxial acceleration sensor (Type-M0053A, Kistler Instruments GmbH) was mounted in the passenger compartment of the vehicle at the airbag control unit. The signals in the x direction alone are of importance for comparison. The other type of acceleration sensor (Early Crash Sensor, Continental) was installed near the crash zone, shown as an additional sensor in Fig. 1. Two such additional sensors were installed, one on the driver side and the other on the passenger side. The positions of both sensors were close to the crash boxes. They measure signals in one direction only (uniaxial) and were mounted on the vehicle structure to measure in x direction. In addition to the acceleration sensors, a pressure-hose sensor was installed in the energy-absorbing foam on the bumper beam. The prototype of the validation sensor manufactured was 130 cm in length (this was limited by the size of the carbon-fiber roll), while the width of the vehicle was 173.5 cm. Hence, it was decided to glue the prototype in such a way that the complete impacting side (driver side) was covered. The prototype was glued on the bumper using superglue (Loctite 401), as shown in Fig. 8b. Since the CFRP-based sensor is glued to the bumper, it measures position and overlap signals in terms of arc length (curved distance) instead of straight distance.

The crash test data were recorded using two data acquisition systems. One system was sampled at a rate of 100 kHz (M=bus Pro Analog module, Messring GmbH), while the other was sampled with a rate of 1000 kHz (LTT24, Labortechnik Tasler GmbH). The systems, along with the power supply box and cable trailing system, were installed in the vehicle trunk, as shown in Fig. 8b. In addition to the data acquisition systems, three high-speed cameras were installed to film from top, left and right views to monitor the crash. The data acquisition systems and the camera system were triggered by a standard trigger sensor used in a crash test. The trigger ensures all the systems have the same zero time (T0) at the first contact of the test vehicle with the rigid barrier.

6.2 Results and discussion

The vehicle crash management structure absorbed the energy of such a critical collision, with some deformation caused to the A-pillar on the driver side. The maximum deformation was measured to be about 95 cm (in the longitudinal direction or x direction as per Society of Automotive Engineers – SAE – norms). It was observed that the raw signals (U_in, I_in and U_m) measured during the crash test contained a noise of 50 Hz. This noise was induced by the DC power supply and the cable system of the crash test facility. A 1-D median filter (“medfilt1” function with n=300 in MATLAB software Mathworks, 2015) was used to remove the noise from the raw signals as shown in Fig. 9a. After filtering the 50 Hz interference, the position and the overlap signals were calculated as explained in the pendulum experiments. Figure 9b describes the reference points used for the crash test. As per the SAE norms, the center of the vehicle width is considered the zero position or reference for position values. For the CFRP-based sensor, this reference or zero position is the end of the sensor at the passenger side. This end is at 40 cm (arc length) from the vehicle center, hence for calculating the relative error, this value is subtracted from the position value. The actual value for calculating the relative error is 17.38 cm, which is the arc length corresponding to the offset 17.35 cm (i.e., the distance between the edge of rigid barrier and vehicle center) and bumper radius of 150 cm (see Sect. 6.3.2 for the calculation of the radius). The position and the overlap values are shown in Fig. 9c and d, respectively. It can be observed from the position signal that the crash event can be detected within 1 ms of the crash. In Fig. 9e frames extracted from the high-speed video at three important time instances are displayed in frame no. 1–3. At about 1 ms after T0, the high resistive layer contacts the low resistive layer at one point, displayed as point 1 in the position signal. As the time progresses, the contact length of the high resistive layer and the low resistive layer increases, which can be observed in both position and overlap signals. The front of the car has a round profile, which is the reason for the increase in overlap with time. At about 6.3 ms, the maximum contact width comes in contact with the barrier. At about 10 ms the partial breakage of the resistive layer starts (the part of the resistive thread in between the contact end points is broken). This was observed by a drop in the current signal. At about 24.95 ms, the bumper is deformed considerably. The passenger-side end of the bumper is detached from the vehicle, causing bending of the sensor. At this time instance, the resistive layer (the unbroken part of the resistive thread) connects with the CFRP layer again and can be observed by the overlap and position signals. One of the contact end points (passenger side) shifts to the other side of the center (caused by the deformation and bending of the bumper). This can also be seen by a drop in the position signal, since it is the midpoint of two contact end points.

Three criteria were considered for comparison of the sensor signals. The first criterion of comparison is the sensor response time for crash detection. A crash event can be identified by a considerable change in the signal, which should be higher than the signal noise. For all the sensors, the time when the threshold of 10 % of the maximum value of the signal from the same sensor is exceeded was considered as the time at which a crash event is detected. The second criterion was whether the sensor can provide additional information about the position of impact. The third criterion was whether the sensor can provide information about the overlap.

Most of the algorithms for activating restraint systems use velocity signals from the acceleration sensor for airbag activation (Kendall and Solomon, 2014). Hence, the acceleration signals from the sensor are filtered with a low-pass filter with a cutoff frequency of 100 Hz and then integrated to get the velocity signals. The derived velocity signals were used for comparison. The central acceleration sensor can only detect a crash event when the value is greater than the threshold (10 % of 20.6 m s⁻¹ is 2.06 m s⁻¹). The acceleration signal from the additional sensor on the crash side (driver side) has higher amplitude compared to the non-crash side (passenger side) sensor. For additional sensors, the maximum of velocity signal of passenger-side sensor was considered for calculating the 10 % threshold for both the sensors. In addition to crash event detection, these sensors can provide information, such as driver-side, central or passenger-side impact.

The pressure-hose sensor also displayed higher pressure on the driver side than passenger side. The maximum pressure for driver-side sensor was about 144 kPa, while for the passenger-side sensor, it was about 25.3 kPa. For the threshold the lower value of the two maximum values was selected (10 % of 25.3 is 2.53 kPa). Since the impact was on the driver side, the pressure signal of the driver-side sensor crosses the threshold (2.53 kPa) earlier (at about 3.40 ms) than the passenger-side sensor (at about 4.48 ms). The position of impact can be calculated using the time difference of the two sensors when the value crosses the threshold pressure, represented as

where the following apply:

• T_right is the response time when the passenger-side pressure sensor crosses the threshold pressure.
  

• T_left is the response time when the driver-side pressure sensor crosses the threshold pressure.
  

• c is the speed of sound in air.
  

The position of the impact calculated with the above equation is with reference to the center of the pressure hose, which is same as the center of the vehicle width (zero position). The pressure-hose sensor cannot provide information on the overlap during the crash.

Table 3Comparison of different sensor types.

^a measured at vehicle tunnel, ^b measured near crash zone (additional sensor) and ^c pressure-hose sensor incorporated in the foam behind the bumper. Please see Fig. 1.

Download Print Version | Download XLSX

Table 3 summarizes the comparison of the sensor signals. It describes the temporal advantage of CFRP-based validation sensor compared to the other sensors. Since the CFRP-based sensor breaks partially before the complete development of the overlap or contact width, the complete overlap value cannot be estimated. But it provides the development of overlap until the sensor breaks partially. The table also shows that CFRP-based sensor can provide all the information which no other sensor can provide.

Figure 10Overlap development: (a) vehicle bumper outline up to 30 ms (red line represents overlap), (b) example frame showing the procedure for drawing vehicle outline, (c) comparison of overlap development of CFRP-based sensor with video analysis results, (d) comparison of overlap development of CFRP-based sensor with predicted with upper and lower limits, and (e) geometrical schematic for predicting overlap development.

Download

6.3 Overlap development

6.3.2 Prediction of overlap development

A vehicle crash against a rigid barrier is considered. The curvature of the bumper can be assumed as a circle with a large radius R, as shown in Fig. 10e. Point C is the center of the assumed circle. B₀ and B₁ are the two contact end points. For the prediction of the overlap development, the lateral movement of the vehicle during the crash is neglected. Hence, the distance of point B₀ from the center line is constant and equal to the position of impact at the first contact point (P_T0 = 17.35 cm). The distance of point B₁ from the center line varies with time, which is dependent on the vehicle deformation x(t) during the crash.

Consider △CA₀B₀ in Fig. 10e, where

$$\begin{array}{}\text{(10)}& \mathit{\alpha }={\sin }^{-\mathrm{1}}\left({\displaystyle \frac{P_{\mathrm{T}\mathrm{0}}}{R}}\right).\end{array}$$

Consider △CA₁B₁, where

$$\begin{array}{}\text{(11)}& \mathit{\beta }\left(t\right)={\cos }^{-\mathrm{1}}\left({\displaystyle \frac{(R\cos \mathit{\alpha }-x(t\left)\right)}{R}}\right).\end{array}$$

The overlap (O(t)) is given by the length of the arc B₀B₁:

$$\begin{array}{ll}{\displaystyle }O\left(t\right)& {\displaystyle }=R\times \left(\mathit{\beta }\right(t)-\mathit{\alpha })\\ \text{(12)}& {\displaystyle }& {\displaystyle }=R\times \left({\cos }^{-\mathrm{1}}\left({\displaystyle \frac{\left(R\cos \mathit{\alpha }-x\left(t\right)\right)}{R}}\right)-{\sin }^{-\mathrm{1}}\left({\displaystyle \frac{P_{\mathrm{T}\mathrm{0}}}{R}}\right)\right).\end{array}$$

The deformation of the vehicle during a crash can be calculated using simplified mass-spring vehicle crash models (Huang, 2002; Lauerer, 2010). An offset vehicle crash can be modeled by a simplified two-mass-spring model, as explained by section frontal offset impact in Huang (2002). The mass of the impacting side was considered to be 60 % of vehicle mass (vehicle mass is 1090 kg), while the stiffness of the impacting side was assumed to be 16 538 N m⁻¹. The remaining 40 % of the vehicle mass was attached to the impacting-side mass by a stiff spring (spring stiffness of 165 380 N m⁻¹). This two-mass-spring model was simulated. The displacement response of the impacting-side mass is the deformation x(t) of the vehicle for the offset vehicle crash scenario. Using Eq. (12) and the deformation calculated from the simplified vehicle model, the overlap development can be predicted. This predicted overlap development can be compared with the measured overlap development using the CFRP-based contact sensor for validation.

For the crash test performed, the radius that fits the test vehicle bumper curvature is 150 cm. Two conditions (time shift and maximum limit) were included in the predicted overlap development. Since the CFRP sensor requires some time after T0 for deformation of the sensor (about 1.2 ms) to determine the position at the first contact point, a time shift of 1.2 ms was added in predicted overlap signal. The overlap development is limited by the maximum contact width, which is the other condition. For the crash test performed, the theoretical maximum overlap is the maximum possible length of arc B₀B₁ (58.1 cm). This can be calculated based on the geometry and considering point B₁ at a distance of 55 cm from the edge of the rigid barrier as shown in Fig. 10e as the maximum limit for the overlap. Figure 10d shows the comparison of the overlap development calculated using equations and the overlap development measured using the CFRP-based sensor. As discussed in Sect. 6.2, considerable deformation and bending of the sensor at 24.95 ms can lead to contact at locations where the sensor bends. Hence, for validation methodology a comparison of the overlap signals until the sensor partially breaks (from 1.2 to 8 ms) should be considered. This is also in line with the time requirements for crash detection and airbag deployment.

As observed by the pendulum experiments, the measured overlap value is always smaller than the actual overlap value. Hence, the lower limit is critical as compared to the upper limit. Two conditions were considered for setting the limit from the predicted value. The first condition is the relative change of the predicted value, which should be considered when the overlap values are low. The other condition is the maximum allowable change from the predicted overlap, which should be considered for large overlap values. For the lower limit at different time instances, the larger value of 50 % of predicted overlap or predicted overlap minus 15 cm was chosen. Similarly for upper limit, the lower value of either 1.25 multiplied by predicted overlap or predicted overlap plus 10 cm was chosen. If for the 90 % of the time instances (between validation time period, 1.2 to 8 ms), the overlap values measured by the CFRP-based sensor are within the upper and lower limit, then the validation is considered positive, otherwise it is considered negative. The conditions for limits are assumed, and these should be optimized based on the requirements of the complete safety system and the possibilities of the restraint system.

Figure 11Generic restraint system activation algorithm.

Download

A rigid barrier test represents a vehicle crash against a rigid wall. Similarly, the equations for a crash with other objects (opponents) like another vehicle, tree, truck, etc. can be calculated using the geometrical simplifications. Forward-looking sensors can also provide information about the geometry of the opponents (Andres et al., 2013; Broßeit et al., 2016; Steinemann et al., 2011). Considering this information along with the other crash parameters (position, velocity, etc.), a particular equation for predicting the overlap development can be selected. Using this methodology, the overlap development can be predicted based on the information of forward-looking sensors.