SE539584C2 - Method and device for determining in-cylinder pressure of a combustion engine - Google Patents

Abstract

24 ABSTRACT Methods and devices for processing a signal generated by asensor (7) adapted to sense pressure variations generated in acylinder (61- 66) of a combustion engine (1) are described. Thesensor is mounted on the combustion engine outside the cylin-der. The sensor signal is low-pass filtered (201) forming at leasta part of an in-cylinder pressure curve, and the in-cylinder pres-sure curve is scaled using a model of the compression in the cyl-inder forming a scaled pressure curve of the at least a part of the in-cylinder pressure curve. (Fig. 2).

Description

The present invention relates to a method and a device for de- termination of in-cylinder pressure of a combustion engine.

BACKGROUND There is a constant aspiration to achieve control of a combustionengine in such a manner that fuel used therein is burned in theengine”s cylinders, while generating a maximum amount of workoutput from the engine and a minimum amount of emissions ofenvironmentally hazardous pollutants. lt is of importance in suchaspiration to have constant knowledge of the combustion en-gine”s operating conditions. One important source of information is the pressure in the cylinder(s) of the combustion engine.

Further, US 8396649 describes a method whereby some in-cylinder pressure data can be reconstructed using a vibration signal from a vibration sensor located outside the cylinder.

SUMMARY OF THE INVENTION lt is an object of the present invention to provide methods anddevices, which at least partly solve the above problems, andwhich are improved in at least some respect in relation to prior art methods and devices.

This object is achieved with the method and the devices as set out in the accompanying claims. ln accordance with one embodiment a method of processing asignal generated by a sensor adapted to sense pressure varia-tions generated in a cylinder of a combustion engine is provided.The sensor is mounted on the engine outside the cylinder. Themethod comprise low-pass filtering the signal from the sensorforming at least a part of an in-cylinder pressure curve. The in-cylinder pressure curve is then scaled using a model of the com-pression in the cylinder forming a scaled pressure curve of the atleast a part of the in-cylinder pressure curve. By first generatingan in-cylinder pressure curve and then scale the thus acquiredcurve it is possible to generate an in-cylinder pressure curve thatwith a low amount of processing represents the in-cylinder pres-sure curve both in terms of its shape and also in terms of the ab- solute pressure values.

The in-cylinder pressure curve thus formed can represent the en-tire in-cylinder pressure curve of a complete working cycle of acylinder or in some applications only a part thereof. The working cycle can be for a four stroke engine or a two-stroke engine. ln accordance with some embodiments the formed in-cylinder pressure curve is phase aligned. ln accordance with some embodiments a part or parts of theformed-in-cylinder pressure curve is replaced by pressure valuesdetermined from a model. The part or parts being replaced cancorrespond to parts of the in-cylinder curve that are determined to comprise noise above a predetermined threshold level. ln accordance with some embodiments the formed in-cylindercurve is smoothened to generate a curve that can be differentiat- ed in each point of the in-cylinder curve. ln accordance with some embodiments the sensor senses signalson the long side of the engine. The sensor can sense signals onthe long side of the engine with the lower temperature comparedto the other long side of the engine. ln accordance with anotherembodiment the sensor is located on the warm side of the en-gine. The signals can typically be a vibration signal or a dis- placement signal. ln accordance with some embodiments the compression model is an adiabatic compression model. ln accordance with some embodiments the sensor senses a vi- bration or a displacement in the engine.

The invention also relates to a device for performing the methodas set out above, and to a motor vehicle comprising such a de- vice.

The invention also relates to a computer program, a computerprogram product, an electronic control device, and a motor vehi- cle.

The invention is not limited to any specific type of combustionengine, but encompasses spark ignited engines as well as com- pression ignited engines, nor to any specific fuel. Non-exhaustive examples comprise fuel in the form of petrol, ethanol, diesel and gas.

Likewise, the invention encompasses combustion engines in-tended for all types of use, such as in industrial applications, incrushing machines and various types of motor vehicles, wheeledmotor vehicles as well as trucks and buses, and boats and crawl- ers or similar vehicles.

Other advantageous features and advantages with the invention are set out in the description below.

BRIEF DESCRIPTION OF THE DRAWINGSBelow are descriptions of example embodiments of the invention, with reference to the enclosed drawings, in which: Fig. 1 is a schematic view illustrating a part of a combustionengine, Fig. 2 is a flow chart illustrating some steps performed whenprocessing a sensor signal, Fig. 3 is a diagram of an electronic control device, Fig. 4 shows a possible placement of a sensor element, Fig. 5 illustrates a sensor signal, cylinder pressure and heat release for a compression cycle, Fig. 6 is similar to Fig. 5 with averaged and filtered signaldata,Fig. 7 illustrates a magnitude spectrum of measured pres- sure and sensed signal Figs. 8 and 9 illustrate a typical adiabatic compression curve, Fig. 10 depicts a filtered and scaled sensor signal, Fig. 11 depicts a signal modelled using pre-compressionmodel, Fig. 12 illustrates a smoothed curve, Fig. 13 illustrates adding of a phase compensation to the sig-nal, Fig. 14 illustrates the use of a post-power model in the signal processing, and Fig. 15 illustrates the final result using an advanced model.DETAILED DESCRIPTION OF EMBODIMENTS Fig. 1 illustrates schematically a combustion engine 1, which en-gine is here arranged in an implied motor vehicle 2, for examplea truck. The engine is equipped with a device 3, indicated with adashed line, adapted to detect operating conditions in the en-gine, and such device has a schematically drawn device 4, which is adapted to detect pressure changes in the cylinder chambers 5 of the combustion engine”s cylinders 61-66, of which there are six in this case, but of which there may be any number.

The device 4 has in this example one sensor element 7 per cyl-inder 61-66, and this is provided outside the associated cylinderchamber 5. The sensor elements are adapted to sense vibrationsor displacements and in particular adapted to sense displace-ments or vibrations in the engine resulting from pressure varia-tions in the cylinders thereof. The terms vibrations/ displacementare used herein to refer to any movement in the engine that canbe sensed by the sensor element 7. The sensors can be piezoresistive or piezo electrical sensors, or other types of vibration ordisplacement sensing elements such as strain measurement de- vice for example a strain gauge.

The device 3 also comprises a unit 9, which may consist of thevehicle”s 2 electronic control device adapted to receive informa-tion about the detected movements from the sensor element(s) 7,and to compare such information, or information calculatedbased on such sensor information, with stored values, and to de-liver measuring values for the state of the engine 1 and/or proc-esses in the engine. Thus, information about the engine”s operat-ing conditions or divergences from these, which suitably providethe bases for control of various components in the combustionengine, such as for example fuel injection, may be obtained based on the sensor elements” 7 detection.

As has been realized pressure changes in the cylinder chamber5, can be sensed by a sensor outside the cylinder chamber to produce high quality signals, which require a simple filtering or processing, to be used when forming an in-cylinder pressure curve or a part thereof.

Fig. 2 shows a flow chart illustrating an embodiment of a methodfor processing a signal generated by a sensor adapted to sensevibrations or displacements generated in a cylinder of a combus-tion engine. The sensor is mounted on the engine outside thecylinder. First, in a step 201, the signal from the sensor is lowpass filtered to form at least a part of an in-cylinder pressurecurve. Then, in a step 203, the in-cylinder pressure curve isscaled using a model of the compression in the cylinder to form-ing a scaled pressure curve of the at least a part of the in-cylinder pressure curve. ln accordance with one exemplary em-bodiment the compression model used in step 203 is an adiabaticcompression model. Additional processing and modelling stepscan also be performed as is indicated by step 205. The order inwhich the steps 201 - 205 are performed can be different in dif-ferent processing implementations for processing the vibrationsignal to form an in-cylinder pressure curve or a part thereof.Processing and modelling steps that can be performed will be exemplified in more detail below.

A computer program code for the implementation of a method ac-cording to the invention is suitably included in a computer pro-gram, loadable into the internal memory of a computer, such asthe internal memory of an electronic control device of a combus-tion engine. Such a computer program is suitably provided via acomputer program product, comprising a non-transitory datastorage medium readable by an electronic control device, which data storage medium has the computer program stored thereon.

Said data storage medium is e.g. an optical data storage mediumin the form of a CD-ROM, a DVD, etc., a magnetic data storagemedium in the form of a hard disk drive, a diskette, a cassette,etc., or a Flash memory or a ROIVI, PROIVI, EPROIVI or EEPROIVI type memory.

Fig. 3 schematically i||ustrates an electronic control device 9comprising execution means 17, such as a central processor unit(CPU), for the execution of computer software. The executionmeans 17 communicates with a memory 18, e.g. a RAIVI memory,via a data bus 19. The control device 9 also comprises a datastorage medium 20, e.g. in the form of a Flash memory or aROIVI, PROIVI, EPROIVI or EEPROIVI type memory. The executionmeans 17 communicates with the data storage means 20 via thedata bus 19. A computer program comprising computer programcode for the implementation of a method according to the inven- tion is stored on the data storage medium 20.

Fig. 4 shows a possible sensor placement. The sensor element 7is here placed on the engine. ln particular the sensor elementcan be placed on a section on the cylinder head. The sensorelements/sensors 7 may be of a suitable type, e.g. piezo resis-tive or piezo electrical elements or optical sensors. The sensorelement may be placed on the engine in an area adjacent to theoutlet of the exhaust channel from a cylinder. For example, itmay be placed on a surface on the engine next to the outlet, onthe engine, of the exhaust channel from a cylinder. The surfacewhere the sensor 7 is placed may be substantially vertical. Thesensor may be arranged to detect vibrations or displacements, which are perpendicular to the movements of the piston. The sensor may also be arranged to detect vibrations or displace-ments, which are perpendicular both in relation to the piston”sdirection of movement and in relation to the engine”s longitudinaldirection. ln one embodiment, the sensor is located on the en-gine”s long side. The sensor may be arranged to detect vibra-tions or displacements in a direction, which is perpendicular in relation to the surface on which it is placed. ln another embodiment (not shown), the sensor element 7 maybe placed in a corresponding manner as when placed on the en-gine at the outlet of the exhaust channel from a cylinder, but in-stead placed in a corresponding location on the engine, at the suction channel”s inlet to a cylinder.

The signal detected by the sensor element 7 may be treated invarious ways as will be exemplified below. The signal from thesensor senses vibrations are low-pass filtered to generate an in-cylinder pressure curve or at least a part thereof. The in-cylinderpressure curve can typically be a continuous curve. The thusformed pressure curve can be used to calculate different valuesat engine control. To enhance the accuracy of the in-cylinderpressure curve, the in-cylinder pressure curve can be processed further in one or more modelling steps and refinement steps.

Below some of such modelling steps and refinement steps aredescribed by way of detailed implementation examples. The in-vention is not limited in any way to the embodiments described,but numerous possible modifications thereof can be envisaged. ln particular steps can be omitted or steps from different em- bodiments can be combined or performed in other sequences than the ones described.

Signal Content Identification An exemplary recurring appearance can be seen in Fig. 5 wherethe sensor signal (knock signal) is plotted together with the cyl-inder pressure. Data is in this example based on a run of1200RPlVl at 100% load. Other setting are of course possible.The sensor data can in accordance with some embodiments beaveraged. For example the data can be averaged over 10 cyclesand Savitzky-Golay smoothed (optimized at Polynomial order 2and frame size 111), the result of the operation is depicted inFig. 6.

The sensor signal can be compared for different operating points to verify if the same appearance could be seen in all modes.

Model development The process in obtaining an estimated pressure model can com-prise at least one of the steps of: filtering out high frequencynoise, adjusting phase shifts, scaling and replacing noisy parts ofthe signal with known physical models or assumptions. Themodel can be based on a high resolution Crank Angle Degree(CAD) signal of for example 0.1 degrees. Also other resolutioncan be used such as 6 degrees resolution. Values in between theacquired data samples can be modelled. The modelled values can for example be generated by a virtual sensor.

To address robustness different models can be used. The models can be combined. Here two models are described. A first model 11 with light signal processing to minimize the model dependenciesand focus on achieving a low average offset of the maximumpressure amplitude relatively to the measurement data. A secondmore advanced model will have heavier signal processing includ-ing phase alignment, post power stroke modelling, high engineload dependencies and more focus on achieving full pressure signal correlation. The two models can also be combined.

Compression Model A compression model can be used for scaling purposes. ln ac-cordance with some exemplary embodiments the compressionmodel is based on the ideal adiabatic equations for compressionof a gas to compensate for engine/load variance as well as pos-sible non-linear Heat transfer losses. For example a Heat trans-fer model based on Woschni/Hohenberger model can be used.The heat transfer model can for example be multiplied by a coef-ficient of in the range of 1 - 10 such as in the range of 1 - 5 inorder to compensate for all modelled losses and the wall tem-perature coefficient can be adjusted depending on engine speedand load to obtain a proper exponential increase during the com-pression stroke. These variables can be tested during the toler-ance analysis in order to investigate the affect these decisions has on the model. ln accordance with one embodiment each step of the process isiteratively calculated based on the thermodynamic first law,where first the number of moles are calculated using the inletmanifold pressure and inlet manifold temperature as initial valuesfor the model as well as the cylinder volume calculation at CAD - 180 degrees, i.e. Bottom Dead Centre BDC prior to Top Dead 12 Centre for combustion. A first estimate of the index, in particularan adiabatic index, is then calculated using an index function. lncase the model is adiabatic the index will be an adiabatic index.The main dependency of this function can be the current tem-perature and lambda. Heat loss from the heat transfer is calcu-lated, which then provides the information needed to calculatethe pressure derivate, the heat transfer is primarily used to com-pensate for the heating loss/gain due to the temperature in thecylinder wall. Temperature, (adiabatic) index and the cylinderpressure can be iteratively calculated up to CAD 160 in order to obtain the full cycle.

Below two models (Light and Advanced) are further exemplified for a typical test implementation.

Pressure model Light filtering The pressure related signal content of the signal from the vibra-tion sensor is of relatively low frequency content compared to thewhole frequency spectrum, the high frequency noise is reducedby applying a low-pass digital filter to the signal. The magnitudespectrum of the relevant frequency range can be seen in Fig. 7which depicts the sensed signal (knock signal) with a measured cylinder pressure (pressure reference).

Different real-time filters such as Butterworth, Elliptic, Chebyfand Cheby2 can be used. Typically there will be a trade-off be-tween filtering too much information near the start of combustion and obtaining a relatively (depending on the operating point) 13 clear signal peak. ln accordance with some embodiments mini-mizing roll-off effects by the real-time filters can be performedwith a Fourier Transform filter. ln accordance with some em-bodiments a 10th order Butterworth filter with a normalized cut-off frequency of about 0.04 can be used. Phase-shift can in someembodiments be avoided by applying zero-phase filtering (for-ward and reverse filtering), hence the signal is then filtered e.g. with two consecutive 10th order Butterworth filters. scaHng The obtained sensor signal can be received in any amplitude. lnaccordance with some embodiments data can be received in or-der of 1-10 V but it could be in other amplitude ranges dependingon the amplifier and settings used as well as the engine speedand load. A pressure model, in particular an adiabatic pressuremodel, can be used to adjust the scaling to the correct level in- dependent of amplifier, settings and the current operating point.

A proper scaling can in accordance with some embodiments beachieved by minimizing the average difference between the fil-tered sensor signal and the pressure model. The minimizationcan for example be performed between two CADs with negativevalues such as -28 CAD and -3 CAD, i.e. during the compressionphase. Here all operating points approximately have the sameappearance (but not necessarily the same amplitude, hence indi- vidual scaling factors are used for each operating point). pre-compression modelThe earlier stage of the compression stroke is cluttered and dis- torted by various noises, to reduce the amount of filtering needed 14 and since the beginning of the compression stroke can be ideal-ized as a compression the model can be replaced between somenegative values such as starting in the range of -150 to -110CAD and ending in the range of- 30 to - 5 CAD. For example -130 CAD to -15 CAD can be used with the scaling model above.This is done on the assumption that the state of the signal in this region coincides with the model used. inlet-exhaust model The opening of the inlet valve will level the cylinder pressure atapproximately manifold pressure, hence the inlet stroke of thesignal can be replaced with an averaged value of the manifoldpressure for the specific operating point to minimize signal filter-ing and smoothing, to reduce the number of needed parametersthe exhaust stroke has been set to the same level. For examplebased on measurements from an experimental campaign, it istrue that the average inlet manifold pressure and exhaust mani-fold pressure are approximately the same, but with availablesensors this assumption can be replaced by the actual exhaustmanifold pressure if required. The replaced inlet region is be-tween some negative CAD values, for example -360 CAD to -130CAD can be used and the replaced exhaust region is betweensome positive CAD values, for example 190 CAD to 360 CAD can be used. smoothing The sensor is typically consistently registering somewhat higherpeaks than the measured pressure curve. ln order to get asmoother peak and minimise the discontinuities between the re- placed pre-compression model and inlet/outlet model the signal can be smoothed. This can in accordance with some embodi-ments be performed using a filter for example using a first orderSavitzky-Golay smoothing filter (for all operating points). Preser-vation of the content surrounding the Start of Combustion (SOC)is significant, hence a second signal can be smoothed with asmaller frame size. A smooth transition can in some embodi-ments be obtained by iteratively comparing the two smoothedsignals and choosing the lower value of them for the regionaround O CAD for example between -5 CAD to +5 CAD or someother range around O CAD.

Pressure model Advanced phase alignment The low pass filtered signal can be phase shifted before scalingin order to raise the power stroke of the signal to the correctlevel. ln accordance with some embodiments the phase shiftmodel used can be a lag compensator, e.g. a second order lagcompensator. ln an exemplary embodiment the position of thepole is p = 1e-12 for all operating points and the position of thezero is varied ranging from z = 4.25e-O4 to z = 3e-04 or someother suitable values with increasing speed and cylinder used.The value can be optimized by comparing measured pressure curve and phase shifted model.

The phase lag compensator will typically only minimize a certainamount of phase indifferences between the signal and the refer-ence pressure signal. ln some examples minimizing further phasedifferences can be done using higher orders of models. However a simple second order model typically works with reasonable re- 16 sults for all operating points without adjusting its parameters too much. scaHngThe filtered sensor signal can be scaled with the same adiabatic compression model as used for the light model above. post-power model The phase adjustment typically does not correct for all irregulari-ties in the power stroke, hence a region during valve opening canbe replaced with a similar model that was used to scale the sig-nal. The region can be around 80 CAD such as between CAD 20to CAD 135. The required initial temperature can be obtainedthrough the thermodynamic first law with total amount of molesobtained from the estimated value in the scaling model, initialpressure and volume taken at around CAD 20 or some CADvalue in that region such as in the range of CAD 10 - 30. lnsteadof compensating for heat transfer losses with a model a simpleoffset based on measurement can be used, the offset can bebased on a sensor measuring the amount of fuel injected and op-timized by comparing the amount injected to the current relativeload (from measurements). The offset for 75% relative load andabove is in accordance with some examples 0.10, 0.07 for 50%,0.03 for 25% and 0 for 0% relative load and motored cycles. Alinear interpolation can be done between the last value, for ex-ample at CAD 135, to the replaced manifold pressure value, forexample at CAD 200. smoothing 17 The signal can be smoothed to adjust scaling issues. This willremove irregularities in the crossovers between models and as-sumptions. For example a 1th Savitzky-Golay smoothing filter with a frame size of 111 can be used.

Combined model A combined model can be developed to obtain better correlationin the power stroke and around the pressure peak. For examplethe light pressure model developed can be used as base with theadvanced pressure model replacing the light model betweensome positive CAD values. The range can start at about 0 - 5CAD and end at about 180 - 200 CAD. For example from CAD2.5 to CAD 195. Hereby a smooth transition is obtained bychoosing the higher value between the two models in the above stated region. smoothing The signal can be smoothed a final time to remove irregularitiesin the crossovers between models. ln accordance with one ex-emplary embodiment a first order Savitzky-Golay smoothing filterwith a frame size of 35 can be used based on measurements fornon-motored cycles and frame size of 51 for motored cycles. Themotored cycles can be indicated by the available sensor measur- ing injected fuel.Test ResultsAdiabatic pressure model The final result of a typical adiabatic compression curve can be seen in Fig. 8 and Fig. 9, where Fig. 9 shows a detail of Fig. 8. ln 18 Figs. 8 and 9 the model curve (Adiabatic model) is compared to the measured pressure (reference pressure).

The adiabatic scaling model can advantageously be individuallycalculated for each operating point prior to scaling to get correct pressure levels from the current sensor readings. filtering The filtered signal can be seen in Fig. 10. The high frequencynoise is removed from the sensor signal and the remaining oscil-lations are related to lower frequency readings. Fig. 10 depictsthe filtered and scaled sensor signal (Filtered and Scaled). Thesignal in Fig. 10 is also scaled using scaling model (here an adiabatic scaling model). pre-compression modelThe result of the replaced pre-compression content is clearlyseen to smooth the curve during the start of compression as seen in Fig. 11 for the pressure model (Pre-compression model). inlet-exhaust modelThe curve in Fig. 11 is the replaced inlet and exhaust contentwhich is very accurately following the measured pressure, the curve (inlet exhaust model) shows the noise replaced. smoothingThe smoothed curve and the final results of the Pressure modellight is seen in Fig. 12, which shows the model (Pressure - model light) compared with the cylinder pressure (Pressure Ref- 19 erence). From the start of the combustion cycle up to near pres- sure peak the model accurately follows the measured pressure.

Pressure model Advanced phase alignment The results of adding phase compensation to the signal is seenin Fig. 13, which shows the model (Pressure - model phasecompensated and scaled) compared with the cylinder pressure(Pressure Reference). The signal has now better correlation withthe measured cylinder pressure during the power and exhaust stroke. scaHng The model is scaled in the same way as the light model using thesame adiabatic scaling model. The results can be seen in Fig.13. post-power model The replaced curve is less noisy and much smoother. The pres-sure peak can in accordance with some embodiments be lifted formost operating points. Smoothing the curve with a larger frameresults in that the increase in peak amplitude is slightly loweredand yields better correlation with the measured pressure trace,the results is seen in Fig. 14, which shows a smoothened curvefor the model curve (Pressure - model postpower and smoothed) compared with the cylinder pressure (Pressure Reference).

Advanced model smoothing A last smoothing can be performed to remove any discontinuitiesin the advanced model. Hence motored cycles can be smoothed.ln accordance with some embodiments a first order Savitzky-Golay smoothing filter with a frame size of 51 and non-motoredcycles are using a frame size of 25 are used. The final result canbe seen in Fig 15, which shows a resulting model curve (Pres-sure - model Advanced) compared with the cylinder pressure (Pressure Reference).

The processing of a vibration signal from a sensor outside a cyl-inder of a combustion engine to generate an in-cylinder pressurecurve is performed using a model. The overall model can involveseveral steps but is nonetheless simple. The filtering can be per-formed with simple real-time Butter-worth filters and performedusing only a single cut-off frequency for all studied operatingpoints, fuels and cylinders. The intake-exhaust model showsminimal difference between the measured pressure at the inletmanifold pressure for the measurements compared herein. ln ac-cordance with some embodiments individual exhaust sensorreadings can be added. The smoothing steps are mainly to mini-mize discontinues between the different models applied and per-formed with relatively few frames. The smoothing performed withthe larger frame is performed mainly to minimize the peak of cer-tain pressure peaks. This can be performed to counter that thesensor signal registers higher peaks than measured with thepressure sensor. The phase alignment is giving an increase in pressure correlation for all operating points in various degrees.

Claims (14)

21 Claims

1. A method of processing a signal generated by a sensor (7)adapted to sense pressure variations generated in a cylinder (61-66) of a combustion engine (1) and where the sensor is mountedon the combustion engine outside the cylinder, characterized by:- low-pass filtering (201) the signal from the sensor forming atleast a part of an in-cylinder pressure curve, and - scaling (203) the in-cylinder pressure curve using a model ofthe compression in the cylinder forming a scaled pressure curve of the at least a part of the in-cylinder pressure curve.

2. The method according to claim 1, wherein the entire in-cylinder pressure curve of a complete working cycle of a cylinder is formed.

3. The method according to any of claims 1 or 2, wherein the formed in-cylinder pressure curve is phase aligned.

4. The method according to any of claims 1 - 3, wherein a part orparts of the formed in-cylinder pressure curve is replaced by pressure values determined from a model.

5. The method according to claim 4, wherein the part or parts be-ing replaced correspond to parts of the in-cylinder curve that aredetermined to comprise noise above a predetermined threshold level. 22

6. The method according to any of claims 4 - 5 wherein theformed in-cylinder curve is smoothened to generate a curve that can be differentiated in each point of the in-cylinder curve.

7. The method according to any of claims 1 - 6, wherein the sen- sor senses signals on the long side of the engine.

8. The method according to any of claims 1 - 7, wherein the compression model is an adiabatic compression model.

9. The method according to any of claims 1 - 8, wherein the sen- sor senses a vibration or a displacement in the engine.

10. A device (3) for processing a signal generated by a sensoradapted to sense pressure changes generated in a cylinder of acombustion engine and where the sensor is mounted on the en-gine outside the cylinder, characterized by: - a low pass filter adapted to low-pass filter the signal from thesensor and adapted to output at least a part of an in-cylinderpressure curve, and - a scaling module adapted to scale the in-cylinder pressurecurve using a model of the compression in the cylinder andadapted to form a scaled pressure curve of the at least a part of the in-cylinder pressure curve.

11. A combustion engine (1), characterised in that it comprises a device according to claim 10. 23

12. A computer program downloadable into an internal memory ofa computer, which computer program comprises a computer pro-gram code adapted to make the computer control the steps ac-cording to any of claims 1 - 9 when said computer program is executed in the computer.

13. A computer program product comprising a non-transitory datastorage medium, which is readable by a computer, and having the computer program code according to c|aim 12 stored thereon.

14. Motor vehicle (2), characterised in that it comprises a com- bustion engine (1) according to c|aim 11.