Method and apparatus for curing a composite article

Abstract

Disclosed is a method of curing a composite article and associated curing apparatus. A heat source is provided for heating the composite article. A temperature related property is detected proximal to the heat source and the heat output is regulated to a predetermined temperature vs. time profile. Heat output vs time data is acquired and functionalised and the curing completion time is determined based on the functionalised heat output vs time data. The method provides for heating a composite article so as to follow a predetermined temperature vs. time profile (i.e. a cure profile) and avoid excessively high temperatures. The required heat output vs time has also been found to be broadly reproducible, and the curing completion time can be more readily determined from functionalised heat output vs time data, for example by identifying reproducible characteristics of the functionalised data.

Claims

1. A method of curing a composite article, comprising; providing a heat source for heating the composite article; detecting a temperature-related property proximal to the heat source; regulating the heat output of the heat source to the composite article, based at least in part on the detected temperature-related property, to a predetermined temperature vs. time profile; acquiring heat output vs time data; functionalising heat output vs time data; and determining a curing completion time based on the functionalised heat output vs time data, wherein the curing completion time is determined by detecting a reproducible characteristic of the functionalised heat output vs. time data corresponding to a local maximum or minimum, steepest gradient, or inflection point in power output vs time data.

2. The method according to claim 1, wherein the heat output is a value of an electrical current or power supplied to the heat source.

3. The method according to claim 1, wherein the functionalising comprises applying one or more mathematical functions to the heat output vs time data.

4. The method according to claim 1, comprising identifying a reproducible characteristic of the functionalised heat output vs time data, and determining the curing completion time based on the reproducible characteristic.

5. The method according to claim 1, comprising identifying a reproducible characteristic of the functionalised heat output vs time data, and predicting the curing completion time based on the reproducible characteristic.

6. The method according to claim 1, wherein the functionalised heat output vs time data is a first derivative of the heat output vs time data.

7. The method according to claim 6, wherein the curing completion time is determined by detecting when the first derivative remains within a threshold range of zero for a predetermined period.

8. The method according to claim 7, wherein the threshold a is within ±0.1, within ±0.01, or within ±0.001 of zero.

9. The method according to claim 6, wherein the curing completion time is determined by detecting a local maximum or minimum in the first derivative of the heat output vs time data.

10. The method according to claim 1, comprising smoothing the heat output vs time data, either before or more after the functionalisation.

11. The method according to claim 10, comprising calculating a rolling average of the heat output vs time data or the functionalised heat output vs time data.

12. The method according to claim 10, comprising acquiring heat output vs time data, or functionalised heat output vs time data, for an evaluation period and then averaging said acquired data.

13. The method according to claim 12, comprising acquiring the first derivative of the heat output vs time data for an evaluation period, averaging the first derivative vs time data over the evaluation period to obtain a smoothed data block, and repeating until a reproducible characteristic is observed in the smoothed data.

14. The method according to claim 13, wherein the reproducible characteristic is a smoothed data block having a first derivative value within a threshold range of zero.

15. The method according to claim 13, wherein the reproducible characteristic is a sequence of smoothed data blocks having first derivative values indicative of a local maximum or minimum.

16. The method according to claim 12, wherein the evaluation period is around 5 minutes or around 9 minutes.

17. The method according to claim 1, comprising providing more than one node, each node comprising a heat source and a temperature sensor.

18. The method according to claim 17, wherein the determining of the curing completion time comprises identifying a reproducible characteristic in the data from at least one of the more than one node.

19. The method according to claim 18, wherein the more than one node includes at least one pair of nodes, and the curing completion time for each of the at least one pair of nodes is determined only when the reproducible characteristic has been identified in the data from both nodes in the at least one pair of nodes.

20. The method according to claim 1, comprising coo in g composite article based on the determined curing completion time.

21. The method of claim 1, wherein the composite article comprises carbon fibre composite.

Description

DESCRIPTION OF THE DRAWINGS

(1) Embodiments of the invention will now be described with reference to the following figures in which:

(2) FIG. 1 shows (a) a schematic cross sectional view and (b)/(c) images of a test curing tool.

(3) FIG. 2 is a schematic of the architecture of a control arrangement that communicates with a heat source and temperature sensor of a curing tool.

(4) FIG. 3 is a schematic cross section of the tool layup of a curing tool.

(5) FIG. 4 is an example cure profile.

(6) FIG. 5 shows power vs time data for a node of a test curing tool during a cure cycle and a re-cure cycle. A difference plot between the two cycles is also shown.

(7) FIG. 6 shows power vs time data for another node of the test curing tool during a cure cycle, together with smoothed first derivative data.

(8) FIG. 7 shows a schematic cross sectional view of the tool layup of a two-sided curing tool.

(9) FIG. 8 shows (a) power vs time data and (b)-(d) smoothed first derivative of the power vs time data, for a selected node of the two sided tool of FIG. 7. Data for the corresponding node of the upper and lower tools is presented.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

(10) A one-sided test tool was used to obtain data on curing of test composite articles, to demonstrate the principles of the invention.

(11) A test curing tool is shown in FIG. 1(a)-(c). The tool 1 is formed of a glass fibre plate 3 (a mould), to act as a mould for curing a test composite article, with heat sources in the form of eleven embedded silicon heater pads 5 (an example of a heat source) and eleven corresponding thin film PT100 resistive temperature sensors 7 (also referred to as resistive temperature detectors, RTDs).

(12) As shown in the schematic cross sectional view of FIG. 1(a), the heater pads 5 and the RTDs 7 are located close to the mould surface 6 of the plate 3. The RTDs 7 are centred above each heater pad 5 to facilitate regulation of the heat output of the respective pad. The selected RTDs provide an accurate and repeatable resistance signal that can be related to temperature (i.e. resistance of the RTD 7 being a temperature-related property) within the target temperature range of 0-200° C. associated with curing of epoxy resin-based carbon fibre composites.

(13) The position of the heat sources is marked on the mould surface 6 (see FIG. 1(b)). FIG. 1(c) shows the tool 1 after a protective PTFE layer has been applied to the mould surface to assist in release of the composite articles during testing.

(14) The plate consists of 16 plies of woven glass fibre infused with an epoxy resin. Glass fibre was selected to prevent electrical shorting of embedded components during manufacture. The heater pads 5 and RTDs 7 are located 1 ply below the heating surface, so that thermal changes within a composite article in the tool 1 can be detected quickly and accurately.

(15) In the embodiment shown, an individual heater pad surface measured 25.4×127 mm and 0.2 mm thick, creating a 127×289.4 mm total surface area with 1 mm gap between adjacent pads. The pads were rated at 1.6 W/cm.sup.2 and used a 240V AC supply.

(16) Each of the heater pads 5 and corresponding RTDs 7 is connected to a respective microcontroller located on the underside of the tool 1 (not visible in FIG. 1). In effect, therefore, the tool 1 comprises a series of nodes 2 formed from a heat source, temperatures sensor and, in this embodiment, a microcontroller.

(17) The architecture of each node 2 is shown in FIG. 2. The control arrangement includes a microcontroller and the relevant software running on a computer processor. The microcontroller 11 is configured to implement Proportional-Integral-Derivative (PID) control to regulate the output of the corresponding heat source 5. The microcontroller 11 schedules the supply of AC power to the heater pad 5 based on feedback from the RTD 7. This is a form of closed loop control, but other means of closed loop control may also be used.

(18) Each node 2 communicates with a processing resource 13, which, in the embodiment shown is in the form of a PC connected via a Universal Serial Bus (USB) interface. The PC is provided with a Graphical User Interface (GUI), by which a user is able to define a cure profile (i.e. a temperature vs time profile) that is then communicated to and locally stored at each individual microcontroller 11.

(19) In use, and as described in further detail below, each microcontroller is operable to regulate the heat output of its heat source, to follow this predetermined cure profile. The cure profile creates a series of temperature set points over time for the microcontroller 11 to drive the heater pads 5 to. The RTD temperature, power consumption and PID values are logged to .csv files, which are stored on the PC.

(20) The microcontrollers used are 8-bit Atmel ATMega 328 with five Analogue-to-Digital Converters 10 (ADC) and five pulse width modulation (PWM) outputs. Although in the embodiment shown each microcontroller 11 is associated with a single temperature sensor 7 and heat source 5, multiple inputs and outputs providing the ability to implement multiple zone controllers with the same microcontroller, in alternative embodiments.

(21) The embedded ADCs are single-ended successive approximations and their channels are multiplexed. They provide 10-bit resolution with an absolute accuracy of 2 Least Significant Bits (LSB).

(22) The power supplied to each of the heater pads 5 is regulated using a zero crossing detector circuit. When the AC sine wave crosses the 0 V axis, a 1 mV pulse is sent from a zero crossing optocoupler or scheduler (ZCS) 12 to the microcontroller 11. The ZCS 12 regulates the AC supply to the heater pad 5 based on a duty cycle determined by the microcontroller 11. The number of pulses can be tracked and the power supply can be turned on and off in for different periods of time, in a similar method to pulse width modulation.

(23) The ADC 10 of the microcontroller 11 is single-ended and utilizes a signal conditioner 14 that consisted of three parts: Constant current source to bias the RTD with a constant current of 1 mA to prevent self-heating which could disrupt the temperature-resistance relationship. Bridge to compensate for wire heating effects connected to the RTD. Amplification stage to amplify the bridge output for better voltage per ADC value (0-1023) on the microcontroller.

(24) The microcontroller's PID compares the current temperature of the RTD to the set point at a given time. Depending on the error margin, the PID sends an appropriate zero-scheduling signal to the heater pad driver circuit. The PID was tuned using the Zeigler-Nichols method using no overshoot parameters. K.sub.i and K.sub.d were set to 0 and K.sub.d was increased until the system oscillates around the set point with constant amplitude of K.sub.u and an oscillation period of T.sub.u.

(25) K.sub.u is the proportional gain coefficient used during the tuning process, and is the value of the proportional coefficient (K.sub.p) that is found to produce a constant oscillation around a given set point when the integral coefficient (K.sub.i) and the derivative coefficient (K.sub.d) are set to zero.

(26) T.sub.u is the oscillation period corresponding to a proportional gain coefficient of K.sub.u

(27) From the values of K.sub.u and T.sub.u, the true K.sub.p, K.sub.i and K.sub.d values were derived using set ratios as described in the Zeigler-Nichols method. This was used to establish baseline PID coefficients (K.sub.p, K.sub.i, K.sub.d) which could then be adjusted by the global controller based on experimental analysis.

(28) Gain scheduling was used to adjust PID gains depending on the current setpoint. For example when the temperature of the plate was similar to the ambient temperature, less aggressive PID values were required to track the set point. In use of the tool 1, as temperature increases, the temperature difference between the plate and the surrounding atmosphere increases, requiring more aggressive PID values to be able to track the set point.

(29) The PID controller 16 was able to maintain the set point temperature within ±0.5° C. This degree of accuracy was observed for all 11 zones or nodes within the tool 1.

(30) Experimental

(31) The following methodology was used for all experiments.

(32) Test Composite Articles

(33) Test composite articles were prepared as follows: Aerospace grade woven pre-preg carbon fibre was used to create 18 ply thick 127×289.4 mm panels. 18 plies were used to create a part with a representative thickness of an aerospace component.

(34) The selected pre-preg carbon fibre required a cure profile comprising a temperature ramp of 3° C./min from ambient to 180° C., followed by a dwell period at this temperature.

(35) Following lay-up, the panels were stored in a laboratory freezer within vacuum bags and were removed 24 hours before curing to allow complete defrosting.

(36) Panel Positioning and Instrumentation

(37) The apparatus setup for curing a test panel 15 (collectively known as the “tool layup”) is shown in FIG. 3.

(38) A panel 15 is centered on the curing tool 1 and encased in two layers of PTFE “peel paper” 17, to prevent bonding to the mould surface 6 of the tool 1.

(39) For the purposes of the test experiments, two RTDs was taped using PTFE coated release film on the top surface of the panels, to provide temperature measurements for the top surface of the plate. These are not shown in the figures.

(40) Cure Lay-up

(41) Curing of the panel requires application of a 99% vacuum, to minimise voids and to compress the carbon fibre lay-up.

(42) A layer of woven polymer breather 18 was placed over the test panel 15 to facilitate air flow, and a flexible vacuum “bag” 20 placed over the test panel and breather so as to contact the mould surface 6 around the periphery of the test panel 15. A tube 21 connects between the vacuum bag 17 and a vacuum pump 19.

(43) Curing

(44) After vacuum was applied, each test panel 15 was cured using the cure profile shown in FIG. 4; i.e. an initial equilibration period at 30° C., a 3° C./min temperature ramp to 180° C., a dwell time of 120 minutes and a temperature ramp of −3° C./min to 30° C., after which the apparatus was switched off and allowed to cool for 4 hours.

(45) Small core samples were removed from each test panel following curing, which were subject DSC analysis to verify that ≥99% curing had taken place, and the cure profile was then repeated for each test panel to obtain comparative data.

(46) The experimental protocol was conducted on six test panels to assess repeatability.

(47) MATLAB (a trademark of MathWorks, Massachusetts, USA) was used for data analysis and visualisation. The data from each RTD was separated and displayed on 11 different graphs, representing the different heater nodes. For the purposes of following, data for individual nodes will be presented.

(48) Results

(49) An example of the power consumption over time of the heat source of a single node during the cure and then re-cure is shown in FIG. 5. The power signals were averaged over 1000 data points using a moving Savitzky-Golay filter (i.e. smoothed data is presented).

(50) For the cure cycle (solid line 23), the power output of the heat source is presented as a percentage of maximum output. Data for heater number 4 is presented in the figure, but comparable data were obtained for other the heaters. Power output initially settles during the equilibration period at 30° C.

(51) During the temperature ramp at 3° C./min, the power steadily increases until peaking at approximately 22% at t=55 mins.

(52) As expected, the power output to maintain the temperature during the dwell period at 180° C. was lower than that required during the temperature ramp.

(53) The power output passes through a local minimum 25 at t≈78 mins and then rises towards a steady rate of around 17.5% at t≈120 mins, where it remains for the rest of the dwell period. The power consumption then during temperature ramp down at −3° C./min.

(54) During the re-curing cycle (dotted line 27), the power output is initially higher the cure profile power consumption and continues to diverge. The power output peaks at 25% at t≈22 min, before falling towards a same steady state of around 17.5% at t≈120 min. Thereafter, power output is as during the cure cycle.

(55) Unlike the power output during the cure cycle, power output during the re-cure cycle did not pass through a local minimum.

(56) DSC analysis on the core samples extracted following the cure cycle confirmed that all tested regions of the test panels were at least 99% cured.

(57) The difference between the power outputs during the cure and the re-cure cycles is shown by line 29 in FIG. 5. This divergence between the power outputs during the cure cycle and re-cure cycle can therefore be attributed to heat liberated by the exothermic epoxy resin curing reaction within the test panel. The required power output to follow the predetermined temperature vs time profile during the re-cure cycle can be regarded as that expected for heating an article having analogous conductivity and heat capacity, taking into account heat losses to the environment for this particular experimental setup.

(58) The initial divergence evident from around t≈22 mins and characterised by a faint shoulder 30 in line 23, is attributable to the onset of the thermally initiated curing reaction. In further experiments (data not shown) a slower temperature ramp rate was used and the onset of curing manifested as a “double minimum” was observed in the power output during curing.

(59) Whilst absolute power output varied, due to differences in heating element power rating, RTD accuracy and increased heat loss near the edges of the test panels (evident in data from nodes 1 and 11 in particular), the difference between cure and re-cure cycles illustrated in FIG. 5 was observed for all nodes in experiments conducted on each of the six test panels.

(60) These data indicate that the exothermic response provides a reproducible characteristics that may be tracked. Two such reproducible characteristics are the approach towards constant power output at t≈120 minutes and the local minimum observed in the test experiments at t≈78 minutes.

(61) Detecting the constant power output of the heat source and/or the approach thereto, is complicated by the differences between absolute power output when this steady state is achieved, due to the aforementioned differences in heat loss from different areas of the tool.

(62) However, the constant power output is characterised by the first derivative of heat output vs time data approaching zero, regardless of these absolute values and can be used to determining the curing completion time.

(63) An example of the first derivative of heat output vs time data (an example of functionalised data) is shown in FIG. 6, by line 31. For comparative purposes, the heat output vs time data from which line 31 was obtained, is shown as line 33. Qualitatively similar results were obtained for all 11 nodes, and indicates that this approach to a first derivative of zero, during a dwell period can be used as a reproducible characteristic for determining the curing completion time.

(64) It has also been observed that the local minimum 35 in the first derivative, at t≈62 minutes (corresponding to the steepest gradient of the heat output vs time 33 between the local maximum 37 and the local minimum 35) is also a suitable reproducible feature of the first derivative of heat output vs time data.

(65) In addition, it has been observed that the time X between the local minimum 35 and the curing completion time (shown at t=120 minutes in this example) is also reproducible, such that detection of the local minimum, in this example at around 62 mins, can be used to predict when the curing completion time will occur. i.e. In this case, X is around 58 minutes.

(66) Two-Sided Test Tool

(67) Experiments were also conducted using a further embodiment of a curing tool 101. Features in common with the tool 1 are provided with like reference numerals, incremented by 100 and 200 for the respective moulds. The tool 101 comprises both an upper mould 103 and a lower mould 203, each of which have an array of 25 nodes 102, 202; as shown schematically in FIG. 7. As previously, each node 102, 202 has a heat source 105, 205 and a temperature sensor 107, 207.

(68) Each mould 103, 203 was constructed generally as described above in relation to the mould 3. However, unlike the mould 3 of the one sided tool 1, both the heat sources 105, 205 and the temperature sensors 107, 207 (as well as their respective microcontrollers 111, 211) of each node 102, 202 was attached to the outer face of the tool the temperature sensor and heater element, to facilitate repair/replacement.

(69) Heat output vs time data for node 17 of the lower mould 203 and data for the upper mould 103 is shown in FIG. 8. Comparable results were obtained from the remaining nodes of both moulds 103, 203.

(70) FIG. 8(a) shows raw heat output vs time data (presented as % of maximum power of the heater pad). FIGS. 8(b)-(d) show smoothed first derivative data, in which the raw data has been averaged for an evaluation period of (b) 1 minute, (c) 5 minutes and (d) 9 minutes. The resulting smoothed data is presented as data blocks each of which has a value of the average extends for an evaluation period in the time dimension. The cure profile was as described above and the start of the dwell period is marked by line 40.

(71) By virtue of the reduced heat losses from the use of both upper and lower moulds 103, 203, the heat output vs. time data (FIG. 8(a)) for node 17 of the lower mould (line 41) and the upper mould (line 43) each equilibrate towards the end of the dwell period at a much lower % power than for the nodes of the one sided tool 1 discussed above. The heat source of node 17 of the lower mould 103 equilibrates at around 2.5% of its maximum power, and the heat source of the node 17 of the upper mould 203 equilibrates at about 5% of its maximum power.

(72) A further consequence of the use of two moulds is that overall heat output form the nodes is lower such that a “double minimum” is observed in the heat output vs time data. A first minimum 45 occurs during the temperature ramp, and arises from heat liberated following the thermal initiation of the curing reaction, and the second local minimum 47 results from heat liberated by the curing reaction during the early stages of the dwell period.

(73) The curing completion time for node 17 was determined from the smoothed first derivative data shown in FIG. 9(d) as follows. The reproducible characteristic was the approach of the first derivative of heat power output towards zero, and curing completion time was determined as when data blocks from both the upper and lower nodes fell within a threshold range of zero.

(74) In this instance, the threshold range T.sub.r was set at ±0.001, which is correlated to ≥99% completion of the curing reaction. In the example shown, the curing completion time was determined at the data block centred at t=100 minutes, marked with reference numeral 48. For manufacturing applications, this would enable this node to be cooled on or after this time. It will be understood that for some applications, a further 1-2 evaluation periods may be allowed to elapse before cooling—for example to allow determination of the curing completion time at other nodes.

(75) In any case, by identifying the curing completion time in this way, the heating may be discontinued far earlier than the t=175 minutes used in the standard cure profile (FIG. 4).

(76) The threshold value of the first derivative depends primarily on the acceptable percentage completion of the curing reaction and was selected to be particularly low in the test experiments, so as to be representative of the requirements for aerospace applications.

(77) The length of the evaluation period will be dependent on experimental conditions, and in particular on the amount of noise in the data.

(78) The evaluation period was determined through analysis of cure cycle temperature and power data from experiments using the same resin system and cure cycle. The power consumption was analysed over the time series and changes in the different cure cycle stages were observed to align with changes in the first order derivative.

(79) Experiments were selected with heater pad and temperature sensor pairs that produced a temperature set point tracking with a standard deviation of less than 1.5° C.

(80) The temperature profiles were used to filter the power consumption profiles by normalising the oscillations in power consumption based on the deviations of temperature from the set point. This was conducted on different time periods until the power consumption standard deviation was reduced to within 5% of rated power for the profile. The optimal time period for the particular resin system described above was found to be 9 mins. Accordingly, the evaluation period can be quantitatively or semi-quantitatively determined from noise or s/n in the temperature data.

(81) The curing completion time for node 17 was also determined from the smoothed first derivative data shown in FIG. 9(c) as follows. The reproducible characteristic was the local minimum of the first derivative of heat power output (corresponding to the local minimum 47 of the raw power vs time data shown in FIG. 8(a)). It was found to be possible to identify the local minimum by a data block from both upper and lower moulds having a first derivative value below a threshold value T.sub.v, in this case of <0.01, wherein the preceding and following data blocks have higher first derivative values. These criteria were satisfied for the data blocks centred at t=47.5 minutes (marked with reference numeral 49).

(82) Thus, by t=58 minutes (reference numeral 51), the local minimum 49 could be identified, and from this the curing completion time predicted, as being a known time period (+50 minutes) from the identified local minimum in the smoothed first derivative.

(83) It should be noted that the local minimum 45 could also be identified at t=46 minutes, by identifying a trend in the first derivative values of successive data blocks, applying a criterion of two successive falls followed by two successive increases.

(84) FIG. 9(a) illustrates that the evaluation period must be selected to be suitable for the amount of noise present in the raw power vs time data. In this example, an evaluation period of 1 minute was not sufficient for the smoothed data to be sufficiently stable for reproducible characteristics to be identified.

(85) It is to be understood that the embodiments of the method and apparatus described above are illustrative examples of the invention and that numerous modifications may be made by one skilled in the art without departing from the scope of the appended claims.