GB2260194A - Engine health monitoring system using crankshaft velocity variation measurements - Google Patents

Abstract

The system determines the rotational velocity of at least one end of the crankshaft in successive samples by counting clock pulses between teeth sensed by magnetic sensors 6, 8. The samples are averaged and processed to provide information relating to the speed variations of the rotation of the crankshaft. The thus extracted information is analysed by an analysis unit 30 and thus an indication of the health of the engine can be obtained. in a preferred embodiment, the diagnostic system comprises a signal conditioner 5 comprising a zero crossing detector and comparator for conditioning the signal from the engine. a sampler 27 and an averager 28. The system can be used to detect unequal fuelling per cylinder, fuel supply faults, sticking valves and cylinder misfiring. <IMAGE>

Description

CRANKSHAFT VELOCITY MEASUREMENTS AND ENGINE HEALTH NONITORING The invention relates to a method of and apparatus for crankshaft velocity measurement and for engine health monitoring and has particular but not exclusive application to engine health monitoring based on crankshaft velocity measurement. The invention may be applied to all type of internal combustion engines.

Medium speed diesel engines are used extensively in a wide range of different applications including power generation, rail traction and automotive applications.

Many ships and large land based vehicles use Paxman Valenta and Ventura diesel engines to generate electrical power and the same type of engines are also used in the propulsion of some ships. In general, an engine is maintained as part of the vehicle's overall maintenance schedule and unpredicted failures can under certain circumstances result in unscheduled maintenance or even decommissioning of the engine, which is clearly undesirable.

Major failures are usually caused by catastrophic events including piston seizure and failure of dampers, crankshafts and con rods. Also, in addition to such major failures, diesel and other internal combustion engines can suffer from component failures which are not in themselves catastrophic but which if not detected can result in a major or catastrophic failure. Broken piston rings cause high blow-by which, if not detected, can result in piston scuffing and seizure. Damper failure due to fluid deterioration or loose securing bolts can lead to crankshaft failure. Ingested debris or broken valves cause piston and liner damage, again with the potential for scuffing and seizure.

One of the most common sources of problems on medium speed diesel engines is the fuel injection equipment, and such faults have included fuel pump seizure and defective injectors. A fuel injection equipment fault will often result in uneven running of an engine, putting additional strain on components which can result in a shortening of the life of the engine.

Currently available engine health monitoring apparatus is limited in it's application, which may in itself contribute to some failures going undetected.

For example crankcase pressure measurement using manometers or pressure transducers has hitherto been the main method for detecting high blow-by. This use of manometers is unreliable because of the difficulties associated with pressurised machinery spaces, and these difficulties have resulted in some seizures due to undetected blow-by. Even when blow-by has been detected by either crankcase pressure or a blow-by meter, it has not been possible to identify specifically the defective cylinder, because of among other things the remoteness of the manometer from the event, i.e. the defective cylinder. Also, sealed oil filled viscous engine dampers have hitherto been monitored by sampling the fluid therein, but fluid sampling is difficult and time consuming and cannot in any case prevent failures due to say a loss of torque in securing bolts.Exhaust temperature is currently sensed by thermocouples, but exhaust temperature thermocouples can be unreliable, reducing the confidence in exhaust temperature readings.

A further problem with some engines is that there is no satisfactory method of measuring cylinder pressures, and even where cylinder pressures can be measured it is not always feasible to monitor continuously all cylinders.

Recent changes in attitudes regarding engine maintenance have resulted in a move away from a time based overhaul approach and toward condition based maintenance. Thus, the above discussed limitations of existing engine health monitoring together with a move toward condition based maintenance has resulted in a real need for an effective condition or health monitoring system. There is therefore a requirement for on-line monitoring of engines.

There has been a considerable amount of interest in using measured variations in crankshaft velocity to provide information on engine state. A large majority of this work has hitherto been aimed at automotive engines with up to 6 cylinders.

A signal generated by the opening and closing of contact breakers can be used to measure the speed of a spark ignition engine, and it has been shown that under idling conditions speed variations can be used to detect a cylinder providing below-normal power. Alternatively, the power contributions made by a given cylinder can be determined by defeating the spark to the cylinder and noting the change in engine speed. Also, inductive probe sensing of teeth on the flywheel of an engine has been widely used to measure crankshaft speed fluctuations. By looking at variations and fluctuations during free acceleration of an engine, defective cylinders have been identified. Furthermore, a measure of engine roughness can be obtained by measuring variations in crankshaft velocity and acceleration on successive engine revolutions.There is a correlation between this roughness and variations in air-fuel ratio and ignition timing.

In a paper entitled "Digital Analyser for Internal Combustion Engines" by C.K. Leong and J.J. Shira, SAE paper no. 820207 it is shown that it is possible to control the air-fuel ratio by monitoring cycle-to-cycle speed variations and to control ignition timing by monitoring in an engine speed signal phase shifts of the harmonic frequency corresponding to the cylinder firing rate.

Freestone and Jenkins disclose in their paper "The Diagnosis of Cylinder Power Faults in Diesel Engines by Flywheel Speed Measurement", T. Mech. E. conference on vehicle condition monitoring and faut diagnosis, March 1985, the detection of cylinder power faults by acquiring measurements of engine speed fluctuations at the engine flywheel and transforming the acquired measurements through a simplified mathematical model of engine dynamics to obtain the effective crankshaft torque variation. This detection is done at less than full speed in order to avoid torsional vibration effects and is accordingly limited in its usefulness.

The interpretation of crankshaft velocities of small engines with typically up to eight cylinders is relatively simple and, despite the shortcomings of the above discussed prior disclosures, the interpretation of signals from small engines is free of additional problems that are encountered on larger engines with many more cylinders. There are many reasons why additional problems are encountered with medium and large engines, one reason being simply that there are a lot more events occurring, e.g. firing of cylinders, for each revolution of the engine. Furthermore, as the number of cylinders increases, the relative contribution of an individual cylinder to the total engine power output decreases. Therefore as the number of cylinders increases it becomes increasingly difficult to detect a small change in power contribution from one particular cylinder.

Another reason why the problem becomes more difficult is that a medium or large engine will have a long crankshaft and long crankshafts are much less stiff, and hence have lower resonant frequencies, than short crankshafts, thereby giving rise to torsional vibration. Also, alternator sets must run at synchronous speeds, and crankshaft torsional vibration is present at these speeds, at both ends of the engine.

The abovementioned method of crankshaft instantaneous speed monitoring for small engines therefore cannot be used for larger engines.

It should be noted that in a four stroke engine a complete engine cycle (induction, compression, power and exhaust) takes two revolutions of the crankshaft. In order to avoid confusion herein the period of events between a cylinder firing and the same cylinder firing again will be referred to as a cycle or engine cycle, this period corresponding to two revolutions of the crankshaft on a four stroke engine.

The firing frequency of a cylinder is therefore half the engine speed and this frequency is regarded as the fundamental frequency and as the first i (half) engine order. Accordingly, an engine order of 1 (one) corresponds to the engine speed, an engine order of 2 (two) corresponds to twice the engine speed or four times the firing frequency, and an engine order of 2i (two and a half) corresponds to two and a half times the engine speed or five times the firing frequency.

It should be apparent to those possessed of the appropriate skills that if cylinder power faults and the like are to be detected in other than small engines by monitoring crankshaft speed fluctuations, a new approach to the problem is required. The present invention aims to provide such an approach.

The present invention aims to provide a method and apparatus for engine health monitoring which meets with at least some of the requirements for on-line monitoring of engine health.

The present invention resides in the realisation that information from instantaneous crankshaft speed measurements can be used in engine health monitoring to increase reliability, minimise initial costs of installation and minimise maintenance both of the engine and of the engine health monitoring equipment.

Crankshaft speed measurements can be processed to provide a large amount of information and this reduces the number of sensors required to be fitted to an engine for the acquisition of data.

According to one aspect of the present invention therefore there is provided an engine health monitoring system in which a processor is arranged to process a signal representing the rotation of at least one end of a crankshaft of an engine to extract therefrom information relating to speed variations of the crankshaft during rotation.

Preferably, the processor comprises a sampler for capturing a sample of the signal from time to time corresponding to at least part of a cycle of the engine and an averager for deriving an average signal from the captured samples.

According to another aspect of the invention there is provided a method of monitoring the dynamic behaviour of an engine crankshaft, the method comprising obtaining a signal representing a parameter associated with the rotation of at least one end of a crankshaft, capturing a signal from time to time corresponding to at least part of the cycle of the engine, deriving an averaged signal from the signal captured from time to time and extracting information relating to torsional vibration of the crankshaft during rotation.

According to a further aspect of the invention there is provided an apparatus for measuring the velocity of an engine crankshaft, the apparatus comprising signal acquiring means mountable to the engine for acquiring a signal representing the rotation of at least one end of the crankshaft, timing means for timing the rotation of the crankshaft between two known angular positions and calculating means for calculating therefrom the velocity of the crankshaft between said angular positions.

The above and further features of the invention are set forth with particularity in the appended claims and together with advantages thereof will become clearer from consideration of the following detailed description of an exemplary embodiment of the invention given with reference to the accompanying drawings.

In the drawings, Figure 1 is a schematic block diagram of an engine and a signal processing system coupled to the engine; Figure 2 is a schematic block diagram of a crankshaft speed fluctuation signal acquisition and conditioning system; Figure 3 is a graph representing a model of torque variation over time of a faulty engine; Figure 4 is a graph representing a model of angular velocity over time of the faulty engine; and Figure 5 is a graph representing a model of angular displacement over time of the faulty engine; Figure 6 is a graphical representation of crankshaft velocity for the condition of (a) underfuelling and (b) overfuelling of a cylinder; Figure 7 is a graphical representation of crankshaft velocity for two different cylinder faults; and Figure 8 is a frequency domain torsional vibration signature of the damper (free end) of a crankshaft.

Referring first to Figure 1 of the accompanying drawings, an engine 1 for driving a load such as an alternator 2 via a crankshaft 3 has associated with it a crankshaft speed fluctuation diagnostic system 4. The diagnostic system 4 comprises a signal conditioner 5 for conditioning signals from sensors, including magnetic sensors and other non-intrusive sensors, mounted to the engine 1. The sensors are mounted to the engine 1 in positions at which they will generate vibration signals and other signals relating to the rotation of the crankshaft 3 and general running of the engine. In practice a suitable location for one magnetic sensor 6 is at or near the engine's damper 7 at the free end of the crankshaft 3.At this location, teeth cut into the damper casing or a fitted toothed wheel will cause the reluctance of a magnetic sensor 6 to vary in accordance with, among other things, the rotational speed variation of that end of the crankshaft 3.

An alternative method of extracting torsional vibration information from the damper/flywheel tooth passing signal is to use a digital timer circuit. With a high frequency clock and triggered counter it is possible to measure the time interval between each tooth passing the magnetic sensor. The string of numbers indicating time intervals can then be plotted out to give a direct representation of the crankshaft velocity.

Another sensor 8 at the engine's flywheel 9, and sensitive to starter ring gear teeth or a fitted toothed wheel (not shown), provides a similar signal representing the flywheel 9 at the other end of the crankshaft 3. A tachometer sensor 10 at a suitable location, eg on the camshaft, provides a reference or once per cycle signal to enable further signal conditioning to be performed. The tachometer sensor is a magnetic or other sensor arranged to sense a preselected engine event. In the case of a two stroke engine the event selected should occur at engine speed, i.e. once per revolution of the crankshaft, and in the case of a four stroke engine the event selected should occur at half engine speed. Such an event can in either case be produced by mounting a suitable sensor to the engine's camshaft.The Tachometer sensor signal is delivered to a once per cycle conditioning circuit 11.

Typically the once per cycle conditioning circuit 11 will provide a once per cycle signal corresponding to the top dead centre (TDC) position of say the first cylinder C1 in one of the banks e.g. bank A (not shown) of the engine 1.

The conditioner 5 acquires and processes signals from the sensors and outputs signals which have been conditioned to facilitate further processing so as to enable the extraction therefrom of information relating to the running of the engine. Signals from the conditioner 5 and from the once per cycle conditioning circuit 11 are supplied to signal analysing equipment 13 for further processing and analysis.

One or more accelerometer 12 mounted on the engine 1 at a known location or locations, e.g. on the top deck between the first and second cylinders C1 and C2 of bank A of the engine 1, provide a linear vibration signal which is input directly to the signal analysing equipment 13. The one or more accelerometers can be used to monitor the effect of combustion faults on linear vibrations, and can be mounted to the top deck of the crankcase. Other accelerometers (not shown) could be mounted in a horizontal axis on the crankcase adjacent to a cylinder main bearings to detect mechanical faults such as excessive bearing clearances.

Any suitably configured signal processing or analysing equipment may be used to analyse the signals output from the signal conditioner 5 but our own MSDA (Mechanical Systems Diagnostic Analyser) is well suited to the task as will be explained in greater detail hereinafter.

Referring now to Figure 2, the sensors 6, 8 each produce an approximately sinusoidal ac output at tooth passing frequency, which signal is shown schematically at 14. The signal 14 is frequency modulated by speed variation of the damper/flywheel and amplitude modulated by the radial component of linear vibration and flywheel orbital motion. The output from each of the sensors is passed to a respective comparator 15 which is arranged such that whenever the signal 14 from the probes 6, 8 passes through zero on a negative slope (i.e. falls below zero as opposed to rising above zero) the output from the comparator 15 goes high. The output from the comparator 15 is shown schematically at 16 in Figure 1.

The comparator output signal 16 is input to a pulse generator 17 which is triggered in response to the signal 16 exceeding a threshold value to produce an output pulse signal 18.

The comparator 15 and the pulse generator 17 thus form a zero crossing detector 19 which produces the train of pulses 18 corresponding to the zero crossing points (-ve going slope) of the magnetic sensor output 5. The function of the zero crossing detector 19 is to minimise the effect of radial motion of the damper/flywheel on the acquired signal. Movement towards or away from the magnetic sensor will produce amplitude modulation of the sensor output but the zero volts crossing point will remain reasonably constant.

The pulse train 18 generated by the zero crossing detector 19 is frequency modulated in the same manner as the input signal 14 from the probe 6 or 8. The pulse train 18 is passed to a frequency to voltage converter 20 comprising a monostable 21 and a set of high pass (HP) and low pass (LP) filters 22. The monostable 11 produces a fixed width pulse 23 of known amplitude every time it is triggered by an input pulse, i.e. a pulse in the pulse train 18. The gaps between the fixed width pulses will vary in proportion to the frequency modulation of the input pulses. The output from the monostable 21 is then averaged to produce an output signal from the filters 22 which output has an amplitude that varies in proportion to the frequency modulation of the input pulse train 18. This averaging is carried out by way of a low pass filter.

In addition the producing an amplitude modulated waveform, the low pass filtering performed by the filters 22 effectively removes unwanted information generated at the frequency at which teeth on the flywheel pass the sensor 6 or 8. The monostable output 23 is also high pass filtered by the filter 12 at low frequency to remove a dc offset corresponding to the steady component of the engine speed. Switchable filters may be used to give maximum flexibility in the signal conditioning process, but in an off-the-shelf system preset filters selected for a given model of engine would be used.

In a typical installation in an engine having a maximum speed of say 1350 rpm and a flywheel starter ring gear with 228 teeth the flywheel tooth passing frequency will be 5130Hz. To remove unwanted information at tooth passing frequency the low pass filter cut off point should be approximately 50% of this frequency, i.e. 2500 Hz approximately. A low pass filter cut off point of 500 Hz and high pass filter cut off of 1 Hz will give a torsional vibration signal bandwidth of 1500 Hz at signals from both the damper and flywheel.

However, the large number of flywheel teeth enables the filter cut off frequency to be increased up to 2.5 kHz if required.

The output waveform from the frequency to voltage converter 20 is proportional to velocity fluctuation of the crankshaft. The output signal from the frequency to voltage converter 20 is supplied to an amplifier 24 and to an integrating amplifier 25. The amplifier 24 simply applies a gain (typically a gain of 100) to the signal output from the filters 22 to facilitate subsequent further processing thereof. The signal from amplifier 24, which is substantially the same as that output from the filters 22 and therefore still represents crankshaft velocity fluctuation, is used in subsequent analysis to detect cylinder power faults, as is explained in greater detail hereinafter. The integrating amplifier 25 integrates the velocity signal to provide a signal representing displacement fluctuation of the crankshaft.

The integrating amplifier 25 also amplifies the signal by a similar gain to that of the amplifier 24. The displacement signal output from the amplifier 25 is used in subsequent run down analysis to measure the amplitudes of displacement fluctuation and can also be used to detect cylinder faults in a similar manner to that of the velocity signal. Of course, it is necessary to ensure that the signal conditioning circuit 5 is accurately calibrated to ensure accurate measurement of engine parameters from the signals output therefrom.

Typically the system 4 will be calibrated so that a torsional vibration amplitude 10 peak to peak will be represented by an output voltage of say 4 volts. System calibration is well within the ability of those skilled in the art and accordingly no further explanation is required or given herein.

Signals output from the amplifiers 24,25 are thus conditioned for input to the signal analysing equipment 13 (see Figure 1) for further analysis.

Signals from the amplifiers 24,25 of the conditioning unit 5 are processed by the processor 13 using the technique of synchronous averaging to provide time domain speed fluctuation signatures. The once per cycle signal generated by the conditioning circuit 11 corresponds to a known position of the engine, for example top dead centre (TDC) of one of the cylinders between the compression and ignition strokes, say. With this information, and knowing the number of cylinders in the engine, it is possible to identify the relative location in time of the same position for every cylinder. Thus, the absolute phase reference produced by the once per cycle pulse enables the position of the other cylinders to be located in the signal average.

Knowledge of cylinder firing positions is essential for identifying the location of any cylinder with a power fault.

The tacho signal, or the once per cycle signal, is used to serve two purposes in the processor 13.

Firstly, it is used by a sampler 27 to divide the signals from the conditioner 5 into portions that correspond exactly to one cycle of the portion of the engine under investigation. Secondly, it is used to synchronise sampling by the sampler 27 of the signal to the rotation of the shaft in order to ensure that the resulting signature is made up of synchronous components relating to the engine. Thus, the once per cycle signal is used to chop the data into discrete sections corresponding to one cycle of the engine. The signal sections are then averaged together by a signal averager 28 such that synchronous components remain but random or non-synchronous components are averaged towards zero.

The averaging process is repeated by the signal averager 28 until the difference between signatures after N and N/2 cycles is less than a preset threshold (1% for example, where N is the numbered sections of data that have been processed). The averager 28 may be either an analogue of a digital circuit.

The crankshaft data is analysed by an analysis unit 30 in the processor 13 in a number of ways in order to present the data in the most suitable form to enable fault conditions to be located and identified. To this end, the incoming signals from the conditioner 5 are synchronously averaged by the averager 28 over an engine cycle to produce stable signals representing velocity or displacement signatures of the damper end of the crankshaft and/or of the flywheel end of the crankshaft.

An average base line signal is derived from incoming data derived from and representing the engine 1 in an initial healthy condition, i.e. a known condition. Data representing the healthy condition signal, i.e the baseline signal, is stored in a database 31. Crankshaft velocity and displacement data is derived by the analysis unit 30 by comparing the incoming signals with the baseline signal and thereafter by way of integration and differentiation of the resulting signals. Frequency domain filtering also is applied to the signals in order to enable selective removal of torsional vibration induced by engine faults.

The information from the analysis unit can be used simply to generate an alarm signal when a fault condition is found to exist. However, as has already been intimated herein, the information can also be used to diagnose the nature of a fault and its location in addition to or instead of simply identifying its existence.

This fault diagnosis process can be considered as having three stages. Firstly, a prediction of the features expected in the crankshaft speed variation signature as a result of different types of faults at different fault locations. Secondly, obtaining the best fit between those predictions and the actual measured features thereby to identify the most likely fault diagnosis and the region of the defect. Thirdly, correlating this result with other engine parameters (exhaust temperatures for example) to maximise the use of all available data in producing a final diagnosis.

In the first stage the model, which as has already been mentioned is a solid body model of the crankshaft system for initial engine health monitoring, is developed initially by theoretical considerations and thereafter by modifying the model to take account of experimental results. This approach leads to an increase in accuracy by accounting for errors that may otherwise be ignored in the theoretical analysis. In the time domain the fundamental shape of the velocity/displacement signatures as a result of solid body effects can be determined for both overfuelling and underfuelling cases. Also, the phase relationship of a feature with respect to an absolute reference point (the once per cycle signal) can be derived for different faults on different cylinders.

An example of a very simple model which gives a first order approximation of the form of the velocity and the displacement signatures (from amplifiers 24 and 25 respectively) generated by solid body effects at the flywheel is as follows. For the purpose of an initial model of underfuelling it will be assumed that the crankshaft/driveshaft is represented by a single inertia I (i.e. there is no torsional vibration); an underfuelling fault causes a step change in torque T for the duration of the firing of the cylinder; the engine governor is controlling the engine such that the mean angular velocity = xm (fault) = xm (no fault) (no fault); and the engine is producing the same power output for the fault and for the no fault case.

Therefore a reduction of fuelling on one cylinder is compensated by an increase in fuelling for the other cylinders.

This variation in torque is represented graphically in Figure 3. In Figure 3 the region 0 to 1 represents the underfuelling of cylinder Al and the region 1 to 2 represents the recovery period in which there is a slight overfuelling of the cylinders.

Torque T is the product of angular acceleration a and inertia I, that is to say T = aI. Therefore the graph in Figure 6 also represents the model of crankshaft angular acceleration a.

The angular velocity X of the crankshaft is the integral of a, i.e.

= 00 + at (1 > Similarly the angular displacement 0 of the crankshaft Similarly is the integral of , i.e.

= owt + ot + iat2 (2) where 6o = o and 0 = X at time t = 0.

Piecewise approximations for regions 0 to 1 and 1 to 2 in Figure 6 gives rise to a piecewise solution to equation (1) as shown in Figure 4 and a piecewise solution to equation (2) as shown in Figure 5.

The plots in Figures 4 and 5 give a first approximation to the variations in angular velocity and angular displacement respectively at the flywheel for the case of the Al cylinder underfuelled. Underfuelling of different cylinders would produce plots of the same form, but with a phase shift dependent on the relative cylinder firing positions. The plots in Figures 4 and 5 represent only crude approximations, and the model requires development from this first approximation to include accurate modelling of the forcing function, friction, inertia, damping effects of the fault on torsional vibration, etc. The model does, however, provide a starting point by explaining the primary solid body effect of fuelling faults.

Thus crankshaft velocity and displacement measurement enable the monitoring of among other things cylinder power balance (or equal fuelling per cylinder) thereby to minimise fuel consumption and prevent uneven thermal/mechanical loading and to maximise component life.

Crankshaft velocity and displacement measurement also enables, for example: automatic cylinder compression; the monitoring of fuel injection equipment faults including an incorrectly adjusted fuel rack; the detection of a worn or seized fuel pump element, or possibly a damaged injector; the detection of broken or sticking air/exhaust valves; and the detection of cylinder misfiring on no-load running.

There is a correlation between the velocity and displacement signatures i.e. signal averages from the signal averager 28 and the position of the cylinders in the engine 1. This correlation enables features of the running of the engine to be identified. For example there is an increase or a decrease in crankshaft velocity following the firing of a cylinder to which too much fuel (an overfuelled cylinder) or too little fuel (an underfuelled cylinder) has been supplied respectively. It is therefore possible to distinguish between underfuelling and overfuelling situations since overfuelling causes a temporary increase in velocity of the crankshaft and underfuelling a decrease in velocity.

The effect of underfuelling one cylinder of a V16 (sixteen cylinder) engine on crankshaft velocity is shown in Figure 3(a) and the effect of overfuelling is shown in Figure 3(b) of the accompanying drawings. The graphs shown in Figure 6 represent crankshaft velocity deviation against time and show the normal crankshaft velocity signature 40 of the engine and an underfuelled velocity signature 41 for the condition of reduced fuel being supplied to cylinder Al, i.e. cylinder number one in bank A and hence the first cylinder in the engine's firing sequence, and an overfuelled velocity signature 42 for the condition of increased fuel supplied to the same cylinder.

It should be noted from Figure 6 that when there is an instantaneous decrease in velocity such as may be caused by underfuelling, the instantaneous velocity of the crankshaft for a period of time thereafter will be higher than it would be under the predefined health condition. This is because the engine governor (not shown) will compensate for an underfuelling fault by subsequently slightly overfuelling the other cylinders in order to keep the mean velocity constant. It should also be noted from Figure 6 that the same applies vice versa to an overfuelling fault.

Figure 7 shows two graphical plots representing different fault conditions introduced at different times to the engine. The graph in Figure 7 represents the overfuelled velocity signature for cylinder Al (line 42) and an overfuelled velocity signature 43 for additional fuel being supplied to cylinder B8 the eight cylinder in bank B. In a V16 engine, these cylinders are in different banks and at opposing ends of the engine but are next to each other in the firing sequence of the engine, i.e. in time.

It will be seen from Figure 7 that there is a clear phase difference between the two signatures 42,43 and that it is therefore possible to distinguish between adjacent cylinders in the firing order. This represents the worst case when it comes to distinguishing between cylinders in a V16 engine and it is therefore possible from a knowledge of the cylinder firing positions to identify the location of any faulty cylinder.

Furthermore, although the two faulty cylinders are at opposite ends of the engine, the amplitudes of the deviations are approximately the same. Therefore, it is possible to distinguish between underfuelling and overfuelling, and also to measure the severity of the cylinder fault, irrespective of the position of the cylinder in which the fault occurs. Also, it should be noted that the difference between the two signals reduces with time toward the end of the signatures.

This is due to the engine governor compensating for the fuelling fault, slightly increasing the fuelling of the remaining cylinders to keep the mean velocity constant.

The dynamic response of the crankshaft 3 can be, and indeed in this embodiment is, modelled as a set of interacting masses, and stiffness representing the behaviour of the crankshaft system. The model will have many modes of vibration, the first mode being that where the crankshaft system vibrates at a frequency having a wavelength that is twice the length of the crankshaft system, i.e. a single vibration node some where between, and two vibration anti-nodes at each end of, the crankshaft system. In this regard it should be noted that the crankshaft system generally comprises the engine per se together with the shaft connecting the engine to a load, and the load. Similarly, the second node corresponds to a vibration situation in which there are two vibration nodes and the third node corresponds to three vibration nodes. Each mode therefore corresponds to a natural frequency of vibration of the crankshaft system and one aim of the processing to be described hereinafter is to match engine models with engine orders of vibration, thereby to find the resonant frequencies of the crankshaft system.

The graphs shown in Figures 6 and 7 represent the change in solid body motion of the crankshaft of the engine in response to power faults, i.e. a 9.5% underfuelling condition, because in reality it is likely that in an underfuelling condition at least some fuel will reach the faulty cylinder during its induction stroke, and an 18.5% overfuelling condition.

Figure 8 shows a frequency spectrum of crankshaft signatures recorded at the damper. The spectrum is derived by the analysis unit 30 applying fast fourier transforms (FFTs) to the incoming signals from the signal averager 28. The damper signature 24 is composed mainly of 1i and 2i order vibration as indicated respectively by the amplitude peaks 45 and 46.

It is beyond the scope of this specification to disclose such modelling in any further detail. However, it has been shown that representational modelling can be used in fault diagnostics.

The purpose of the above described modelling is to provide reference data representing different class and forms of engine failure. The most important features associated with a fault type are selected from the results of simulations and experimental data. These features are then used in a fault classification process to separate the different types and locations of faults.

Fault classification may be best carried out in an engine health monitoring system, utilising information from other engine parameters such as exhaust temperatures. Depending on requirements, the classification process could be very simple or use more sophisticated techniques. The latter will be required if the monitoring system is to be able to diagnose multiple faults, rather-than give a simple warning without diagnosis.

In addition to using crankshaft speed variation data in order to monitor engine performance it is possible to use data acquired from other sources to the same end. For example exhaust branch temperatures can provide valuable information about engine performance.

For example data defining mean exhaust temperature, temperature scatter (range between maximum and minimum values) and deviation from the mean temperature can be derived from the incoming data by the analysis unit 30.

Such figures can then be compared with, or otherwise analysed with reference to, predetermined information in the data base 3 to analyse and identify fault conditions in the engine 1. Changes in exhaust temperature give a clear indication of cylinder power related faults.

However, it is not desirable to rely solely on the monitoring of exhaust temperatures. This is because for example a separate thermocouple is required for each cylinder of the engine and because of the environment in an exhaust the thermocouples would require regular maintenance and replacement. Furthermore, exhaust temperatures can be influenced by several factors, in addition to cylinder power. For example, an increase in exhaust temperature may be caused by a burnt exhaust valve, a sticking valve, retarded injection timing, insufficient excess air, a fouled intercooler, and so on. It is not always possible to distinguish between several of these factors from a knowledge of exhaust temperature alone. Nevertheless, the use of exhaust temperature data in combination with crankshaft torsional signatures provides for an engine monitoring system of improved capability.

It is also possible to monitor the conditioned damper 7 by monitoring torsional vibration levels at the damper end of the crankshaft 3. Torsional vibration is a result of the excitation of crankshaft natural frequencies (modes) by different forcing functions (engine orders) generated by the firing sequence. As the engine 1 is a four stroke engine, one firing cycle is completed every two revolutions, and hence crankshaft half-orders are present. The resulting torsional vibration is the total effect of the excitation of the different crankshaft modes by multiple engine orders.

In the case of a 16 cylinder Valenta engine with an alternator and supercharger drive, it can be shown that for the important first crankshaft mode torsional vibration amplitudes at the flywheel should be about 30% of those at the damper, i.e. torsional vibration is present at both locations. This is not necessarily always the case, but in general the torsional vibration at the flywheel will usually be lower than that of the damper.

One of the most reliable methods of using torsional vibration to monitor the damper condition is to select an engine order which passes through a crankshaft resonance in the engine operating speed range and track this order during a run-down. The damper condition can be inferred from a knowledge of the frequency of, and vibration amplitude at, the resonance. Thus tracking engine orders during a run-down of the engine is used to monitor the condition of the damper 7. Changes in the damper condition can be detected from changes in position and magnitude of an order having a large amplitude (for example the 2i order peak 46 in Figure 5) over a period of time.

The above described embodiment of the invention uses crankshaft velocity measurement to monitor the health of an engine. The embodiment provides for increased engine protection by detecting faults and enabling intervention to prevent an engine operating to serious damage or destruction. For example, the detection of increased blow-by before a piston seizure, and damper deterioration before crankshaft failure can be detected.

The embodiment also provides for higher operational availability by allowing maintenance intervention to keep the engine running. Furthermore, the embodiment provides for an improved discrimination between faults by providing instantaneous crankshaft speed data which can be combined with other information, such as cylinder exhaust branch temperatures, and in so far as we are aware such information has not previously been made available for combination. This in turn can reduce maintenance workload or repair time by improving the ability to discriminate between faults and identify an affecting cylinder. Assuming that the appropriate maintenance management system is in place, thus can lead to improved maintenance scheduling by an earlier knowledge of faults as they develop.

A thus improved knowledge of engine performance leads to a better control of fuel consumption.

The embodiment offers a simple and rugged solution to engine health monitoring, requiring no engine modifications which is therefore suitable for retrofitting to existing engines. The system can be contained in a small box and is capable of both local and remote operation.

Having thus described the present invention by reference to a preferred embodiment it is to be well understood that the embodiment in question is exemplary only and that modifications and variations such as will occur to those possessed of appropriate knowledge and skills may be made without departure from the spirit and scope of the invention as set forth in the appended claims and equivalents thereof.

Claims (32)

CLAIMS:

1. An engine health monitoring system in which a processor is arranged to process a signal representing the rotation of at least one end of a crankshaft of an engine to extract therefrom information relating to the speed variation of the rotation of the crankshaft.

2. An engine health monitoring system as claimed in claim 1, wherein the processor comprises a sampler for capturing a sample of the signal from time to time corresponding to at least part of a cycle of the engine and an averager for deriving an average signal from the captured samples.

3. An engine health monitoring system as claimed in claim 2, wherein the processor further comprises an analysis unit for analysing the signal from the averager.

4. An engine health monitoring system as claimed in claim 3, wherein the analysis unit is arranged to compare the signal from the averager with a previously derived signal representing the engine in a known state thereby to enable identification of a fault condition in the engine.

5. An engine health monitoring system as claimed in claim 3 or 4, wherein the analysis unit is arranged to compare a previously derived baseline signal with the averager signal and to analyse the resulting signal to enable identification of a fault condition.

6. An engine health monitoring system as claimed in any of claims 3 to 5, wherein the analysis unit is arranged to integrate or differentiate the averager signal in order to derive velocity or displacement data respectively from the signal.

7. An engine health monitoring system as claimed in any of claims 3 to 6, wherein the analysis unit is arranged to compare the averager signals or signals derived therefrom with predefined model signals stored in a data store.

8. An engine health monitoring system as claimed in any preceding claim, wherein a sensor is mountable to the engine in the vicinity of the flywheel in order to provide a signal representing the rotation of the flywheel end of the crankshaft.

9. An engine health monitoring system as claimed in any preceding claim, wherein a sensor is mountable to the engine in the vicinity of the free end of the crankshaft in order to provide a signal representing the rotation of that end of the crankshaft.

10. An engine health monitoring system as claimed in claim 6 or 7 wherein the or each sensor cooperates with a respective toothed disc mounted to the crankshaft.

11. An engine health monitoring system as claimed in any preceding claim, wherein a sensor is mountable to the engine to provide a signal relating to the rotation of the engine.

12. An engine health monitoring system as claimed in any preceding claim, wherein a conditioner is arranged to condition the signal or signals from the engine in order to facilitate processing by the processor.

13. An engine health monitoring system as claimed in claim 12 wherein the conditioner comprises a zero crossing detector for detecting when a signal from the engine crosses a zero value and outputting a signal in response thereto.

14. An engine health monitoring system as claimed in claim 13 wherein the zero detector comprises a comparator for comparing the signal or signals from the engine with a respective reference level or levels and a pulse generator responsive to the comparator for generating a pulse representing a zero crossing.

15. An engine health monitoring system as claimed in claims 13 or 14 wherein the comparator further comprises a frequency to voltage converter responsive to the frequency of the signal output from the zero crossing detector for outputting a signal having a voltage representative of the said frequency.

16. An engine health monitoring system as claimed in claim 15, wherein the frequency to voltage converter comprises a monostable responsive to the output signal from the zero crossing detector for generating a pulse signal having a constant mark and variable space dependent on the frequency of the zero crossing of the signal from the engine.

17. An engine health monitoring system as claimed in claim 16 wherein the frequency to voltage converter further comprises a filter or filters for converting the pulse signal from the monostable into a signal representing torsional vibration velocity of the crankshaft.

18. An engine health monitoring system as claimed in any of claims 12 to 17 wherein the conditioner further comprises an amplifier for amplifying signals for delivery to the processor.

19. An engine health monitoring system as claimed in claim 18 wherein the amplifier is adapted to integrate a crankshaft velocity signal derived by the conditioner, thereby to produce a signal representing torsional vibration displacement of the crankshaft.

20. A method of monitoring the dynamic behaviour of an engine crankshaft, the method comprising obtaining a signal representing a parameter associated with the rotation of at least one end of a crankshaft and extracting information relating to speed variation of the crankshaft during rotation.

21. A method as claimed in claim 20, further comprising capturing a signal from time to time corresponding to at least part of the cycle of the engine, deriving an averaged signal from the signal captured from time to time, the information being extracted from the averaged signal.

22. A method as claimed in claim 20 or 21, further comprising comparing the averaged signal with a previously derived signal representing a parameter associated with a known-dynamic behaviour of the crankshaft.

23. A method as claimed in any of claims 20 to 22, further comprising subtracting a previously derived baseline signal from the averaged signal to produce a fault signal containing substantially only parameters associated with a fault condition.

24. A method as claimed in claim 23 further comprising analysing the fault signal to identify a fault in the dynamic behaviour of the crankshaft.

25. A method as claimed in any of claims 20 to 24 further comprising obtaining a signal representing the rotation of the other end of the crankshaft.

26. A method as claimed in any of claims 20 to 25 further comprising obtaining a tacho signal from the engine and deriving therefrom a signal relating to the cycle of the engine for use in the capturing of the signal from the engine from time to time.

27. An apparatus for measuring the velocity of an engine crankshaft, the apparatus comprising signal acquiring means mountable to the engine for acquiring a signal representing the rotation of at least one end of the crankshaft, timing means between two known angular positions and calculating means for calculating therefrom the velocity of the crankshaft between said angular positions.

28. An apparatus as claimed in claim 27 wherein the calculating means is arranged to calculate repeatedly the velocity of the crankshaft so as to provide velocity data for at least one engine cycle.

29. An apparatus as claimed in claim 27 or 28, further comprising an integrating means or integrating the velocity data to produce data representing the displacement of the crankshaft.

30. An apparatus as claimed in claim 28 or 29, wherein the signal acquiring means comprises a magnetic sensor responsive to the rotation of a toothed member associated with the crankshaft, and the timing means comprises a zero crossing detector for detecting the signal from the signal acquiring means crosses a zero level and means responsive to the zero crossing detector for providing a signal representing the time interval between detected zero crossings.

31. An engine health monitoring system substantially as hereinbefore described with reference to the accompanying drawings.

32. A method of monitoring the dynamic behaviour of an engine crankshaft the method being substantially as hereinbefore described with reference to the accompanying drawings.