System and method for identifying combustion anomalies of internal combustion engines by means of indirect measurement

Abstract

The invention relates to a computer-implemented method (100) for identifying combustion anomalies by indirectly measuring, by means of a classifier (2), status data of an internal combustion engine (1, 1a, 1b, 1c) to be analyzed, in which method: operating data of the internal combustion engine (1, 1a, 1b, 1c) to be analyzed are acquired; at least one feature is generated for the classifier (2) by processing at least a portion of the operating data by means of mathematical operations and / or by selecting a data range from the operating data; and status data of the internal combustion engine (1, 1a, 1b, 1c) to be analyzed are determined by applying the classifier (2) to the generated features.

Description

[0001] System and method for detecting combustion anomalies in internal combustion engines by means of an indirect measurement

[0002] The invention relates to a system and computer-implemented method for the indirect measurement of state data of an internal combustion engine to be analyzed, in particular the detection of combustion anomalies by means of a classifier. Furthermore, the invention relates to a system and method for training such a classifier.

[0003] The safe and efficient operation of an internal combustion engine is crucial for the functionality of modern vehicles or drive systems.

[0004] However, current methods for detecting combustion anomalies are often complex, expensive, or have limited accuracy. For example, direct measurement methods such as monitoring temperature and pressure readings, especially in the cylinder, can be difficult to implement, particularly in situations where access to the combustion engine is restricted.

[0005] EP 3 693 588 A1 discloses a knock sensor that detects the vibration of an engine block and a pressure sensor that detects the pressure in a combustion chamber. A value representing the knock intensity is derived from the output values ​​of the pressure sensor. Weights of a neural network are learned using a value representing the engine block vibration detected by the knock sensor as input to the neural network and the detected value representing the knock intensity as training data. The value representing the knock intensity is then estimated from the output values ​​of the knock sensor using the learned neural network.

[0006] US Patent 5,093,792 A discloses a device for predicting and differentiating, based on cylinder pressure and using a three-layer neural network, whether misfires, knocking, and the like will occur before they actually happen. The cylinder pressure, detected by a cylinder pressure sensor, is sampled and fed into each element of the input layer. The signal is then modulated according to the strength (weight) of the connection between each element and passed to the hidden and output layers. The magnitude of the signal from the elements of the output layer represents the prediction and differentiation results. The weight is learned and determined by a backpropagation process.

[0007] It is an object of the invention to provide an improved system and method for detecting combustion anomalies. In particular, it is an object of the invention to make the detection of combustion anomalies during normal operation, especially in a production engine, more reliable and to differentiate between different combustion anomalies more accurately.

[0008] A first aspect of the invention relates to a computer-implemented method for the indirect measurement of state data of an internal combustion engine to be analyzed, in particular for the detection of combustion anomalies by means of a classifier, comprising the following steps:

[0009] • Recording operating data of the internal combustion engine to be analyzed, wherein the operating data includes value profiles of measurement parameters and characterizes operating behavior;

[0010] • Generating at least one feature for the classifier by processing at least a portion of the operational data using mathematical operations and / or by selecting a data range from the operational data; and

[0011] • Determining state data of the internal combustion engine to be analyzed by applying the classifier to the generated features; wherein the operating data comprise as measurement parameters a rotation period with respect to a defined angular range of a drive shaft and an engine speed of the internal combustion engine to be analyzed, on the basis of which a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the internal combustion engine to be analyzed are generated as features, and wherein the state data characterize physical states, in particular a pressure and / or a heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engine to be analyzed.

[0012] A second aspect of the invention relates to a computer-implemented method for generating a classifier for determining state data of an internal combustion engine, in particular for detecting combustion anomalies of the internal combustion engine by training a classification algorithm, comprising the following steps:

[0013] • Acquisition of operating data of a plurality of internal combustion engines of a specific type of internal combustion engine, wherein the operating data includes value profiles of measurement parameters and characterizes behavior during operation, and of state data of the respective internal combustion engine during operation;

[0014] • Generating features for a classifier by processing at least a portion of the operational data using mathematical operations and / or by selecting a data range from the operational data; and

[0015] • Training the classification algorithm using the generated features and their respective associated state data, thereby generating the classifier; wherein the operating data comprise, as measurement parameters, a rotation period with respect to a defined angular range of a drive shaft and an engine speed of the internal combustion engines, based on which, as features, a change in the rotation period with respect to the defined angular range, a mean engine speed, and an engine torque of the respective internal combustion engine are generated, and wherein the state data characterize physical states, in particular a pressure and / or a heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engines. A third aspect of the invention relates to a system for the indirect measurement of state data of an internal combustion engine to be analyzed, comprising:

[0016] • a rotary encoder for recording operating data of the internal combustion engine to be analyzed, wherein the operating data include a value profile of at least one rotation period with respect to a defined angular range of a drive shaft and a value profile of an engine speed of the internal combustion engine to be analyzed as measurement parameters and characterize an operating behavior;

[0017] • a data processing device with means for generating features for a classifier by processing at least a part of the operating data by means of mathematical operations and / or by selecting a data range from the operating data, wherein the features generated are a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the internal combustion engine to be analyzed, and with means for determining state data of the internal combustion engine to be analyzed by applying the classifier to the generated features, wherein the state data characterize physical states, in particular a pressure and / or heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engine to be analyzed.

[0018] A fourth aspect of the invention relates to a system for generating a classifier for the indirect measurement of state data of an internal combustion engine by training a classification algorithm, comprising:

[0019] • a rotary encoder for acquiring operating data of a plurality of internal combustion engines of a type, wherein the operating data comprise a value profile of a rotation period with respect to a defined angular range of a drive shaft and a value profile of an engine speed of the respective internal combustion engine as time-resolved measurement parameters and characterize an operating behavior; • a sensor for acquiring state data of the respective internal combustion engine during operation, wherein the state data characterize physical states, in particular a pressure and / or heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engines; and

[0020] • a data processing device with means for generating features for a classifier by processing at least a part of the operating data by means of mathematical operations and / or by selecting a data range from the operating data, wherein the features generated are a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the internal combustion engines, and with means for training the classification algorithm using the generated features and the respective associated state data, whereby the classifier is generated.

[0021] A fifth aspect of the invention relates to a computer-implemented classifier for the indirect measurement of state data relating to combustion anomalies of an internal combustion engine, wherein the classifier is generated by training a classification algorithm, wherein the classification algorithm was configured by the following steps, which are performed for each training input of a plurality of training inputs:

[0022] • Recording operating data of an internal combustion engine of a specific type of internal combustion engine, wherein the operating data includes value profiles of measurement parameters and characterizes behavior during operation, and state data of the respective internal combustion engine during operation;

[0023] • Generating features for a classifier by processing at least a part of the operating data using mathematical operations and / or by selecting a data range from the operating data; and • Training the classification algorithm using the generated features and the respective associated state data, whereby the classifier is generated; wherein the operating data comprise, as measurement parameters, a rotation period with respect to a defined angular range of a drive shaft and an engine speed of the combustion chamber of the respective internal combustion engine, on the basis of which, as features, a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the respective internal combustion engine are generated, and wherein the state data characterize physical states, in particular a pressure and / or a heat and / or an internal engine torque, of at least one combustion chamber of the respective internal combustion engine.

[0024] Preferably, the internal engine torque is a function of cylinder pressure and piston position or crankshaft angle.

[0025] An internal combustion engine within the meaning of the present disclosure is preferably a reciprocating piston engine or a Wankel engine.

[0026] A combustion anomaly within the meaning of the present disclosure is preferably an unwanted early or late pre-ignition, a delayed ignition, a misfire or knocking.

[0027] For the purposes of this disclosure, data acquisition preferably involves reading measured operating data via a data interface. Alternatively or additionally, data acquisition includes determining a measurement signal using a sensor and / or post-processing a measurement signal to generate the operating data.

[0028] A mathematical operation within the meaning of the present disclosure preferably includes statistics, in particular mean, minimum, maximum or standard deviation, or aggregation, addition, subtraction, multiplication, division, differentiation and / or integration over certain durations and / or multivariate integrals in which measurement parameters are multiplied and then integrated.

[0029] A class of internal combustion engines within the meaning of this disclosure is preferably a set of technical devices which are identical in their essential features and are therefore preferably of identical construction. Preferably, the essential components of the internal combustion engines of a class are identical in construction. A specific internal combustion engine is thus preferably an implementation of the class of internal combustion engines. In particular, internal combustion engines of a class differ by tolerances, especially manufacturing tolerances and / or aging or wear effects.

[0030] An indirect measurement within the meaning of the present disclosure preferably involves determining a desired measured value of a parameter using other available information from a system to be measured. More preferably, the desired measured value of a parameter is determined from at least one other physical parameter.

[0031] State data within the meaning of the present disclosure preferably include a cylinder pressure, an indicator parameter of the cylinder pressure, in particular an indicated mean effective pressure (IMEP) and / or a crank angle at which half or 50% of the fuel is burned (MFB50), a peak combustion pressure and / or a heat release rate.

[0032] Peak Firing Pressure (PFP), as defined in this disclosure, preferably describes the maximum pressure in the combustion chamber during the power stroke. It is the highest pressure reached when the air-fuel mixture burns and is a critical parameter for engine performance and durability. A high PFP can lead to increased engine power and efficiency, but if it is too high, it can cause engine knocking, increased wear, and potential damage to engine components. Controlling the PFP is important to optimize engine operation and achieve a balance between performance, fuel efficiency, and durability without mechanically overloading the engine.

[0033] A heat release rate (RoHR) in internal combustion engines, as defined in the present disclosure, is preferably the rate at which energy is released from the combustion of the air-fuel mixture in the combustion chamber of the engine. It is typically measured in joules per degree of crank angle (J / °CA) or watts (W) and provides important information about the combustion process, including its efficiency and the timing of combustion.

[0034] RoHR is an important parameter for several reasons:

[0035] • Combustion efficiency: This helps to understand how effectively the fuel is burned.

[0036] • Performance tuning: By analyzing the RoHR, the timing and amount of fuel injection can be optimized to improve engine performance.

[0037] • Emission control: Controlling RoHR can contribute to reducing harmful emissions.

[0038] • Engine knocking: Monitoring and controlling the RoHR can prevent engine knocking, which can cause damage.

[0039] Understanding and controlling RoHR are important to achieve optimal engine performance, fuel efficiency, and emission control.

[0040] A means as defined in this disclosure can be configured as hardware and / or software and, in particular, comprise a processing unit, preferably a microprocessor (CPU), preferably connected to a memory and / or bus system via data or signals, and / or one or more programs or program modules. In other words, the means can be a hardware and / or software unit comprising a processing unit, preferably a microprocessor (CPU), and / or one or more programs or program modules. The processing unit or microprocessor can be connected to a memory and / or bus system via data or signals. The CPU can be configured to execute instructions implemented as a program stored in a memory system, to acquire input signals from a data bus, and / or to output signals to a data bus.A storage system can comprise one or more, in particular different, storage media, especially optical, magnetic, solid-state and / or other non-volatile media. The program can be designed in such a way that it embodies or is capable of executing the procedures described herein, so that the CPU can execute the steps of such procedures and thus, in particular, analyze at least one technical device or train a classification algorithm.

[0041] The term "means" as used herein encompasses all structures, materials, or actions set forth herein, as well as all equivalents thereof. Furthermore, the structures, materials, or actions and their equivalents include everything described in the abstract, the brief description of the figures, the detailed description, the summary, and the claims themselves. A system and / or its means may preferably take the form of a purely hardware variant, a purely software variant (including firmware, resident software, microcode, etc.), or a combination of software and hardware aspects, generally referred to as a "circuit," "module," or "system." Any combination of one or more computer-readable media may be used. The computer-readable medium may be a computer-readable signaling medium or a computer-readable storage medium.

[0042] The systems and methods according to the present disclosure can preferably be implemented in conjunction with a suitably configured computer, a programmed microprocessor or microcontroller and one or more peripheral integrated circuit elements, an ASIC or other integrated circuit, a digital signal processor, a hard-wired electronic or logic circuit, such as a circuit with discrete elements, a programmable logic device or gate arrangement, such as a programmable logic device (PLD), a programmable logic array (PLA), a field-programmable gate array (FPGA), a programmable logic arrangement (PAL), or a comparable means.In general, any device or means capable of implementing the methodology presented herein may be used to implement the various aspects of this disclosure. Exemplary hardware includes computers, handheld devices, telephones (e.g., cellular, internet-enabled, digital, analog, hybrid, and others), and other hardware known in engineering. Some of these devices include processors (e.g., a single or multiple microprocessors), memory, non-volatile memory, input devices, and output devices. Furthermore, alternative software implementations, including but not limited to distributed processing or distributed processing of components / objects, parallel processing, or processing by virtual machines, may be developed to implement the procedures described herein.

[0043] The invention is based on the analysis of operating data of the internal combustion engine to be analyzed and uses a classifier to convert the generated features into state data.

[0044] The method is characterized by the fact that no direct measurement of combustion or combustion process parameters, in particular temperature or pressure in the combustion chamber, is required. Instead, operating data of the internal combustion engine are acquired and processed to create features that reflect the engine's condition. This enables diagnostics of the internal combustion engine during normal operation, especially in the case of a production engine, which is otherwise only available during testing, particularly on a test bench. The position of the 50% mass conversion point (MFB50) is used as a reference variable for combustion control.

[0045] These characteristics can then be evaluated by a classifier to determine the condition data of the combustion engine. This indirect measurement allows the method to be used even in situations where direct measurement methods are not possible or pose difficulties.

[0046] The present invention offers the advantages of a non-invasive measurement method that can operate in real time, but does not require pressure measurement in the cylinder. The invention can be used in various industrial sectors, such as the automotive industry or in other industrial applications.

[0047] The rotation period of an internal combustion engine's drive shaft, measured through a defined angular range, is a parameter that can be directly measured at the shaft itself. This parameter indicates how quickly the shaft has rotated through a specific angular range. Its advantage lies in its ability to be directly determined for relatively small angular ranges using state-of-the-art measurement methods, such as a rotary encoder. This provides numerous individual measurements for each revolution of the drive shaft, each indicating the corresponding rotation period. Because the rotation period is measured for the same angular range, it is immediately meaningful and does not require conversion into a quotient. This avoids information loss due to calculations, such as rounding.Furthermore, the calculation of this rotation period for a specific angular range is a basic parameter in an engine control unit, which is calculated directly from the value of a timer, which in turn is derived from the high-precision clock frequency of a processor of the engine control unit.

[0048] In an advantageous embodiment, the method, according to the first aspect, further comprises the following step:

[0049] Controlling the internal combustion engine based on specific condition data.

[0050] By considering the engine's operating conditions when controlling it, critical engine states can be detected or predicted more effectively and earlier. Therefore, the engine can be operated within optimal ranges that are even closer to these critical states.

[0051] In an advantageous embodiment of the methods according to the first or second aspect, the internal combustion engine is a hydrogen combustion engine. A precise analysis of the combustion process is particularly important for hydrogen engines: hydrogen is very flammable compared to other fuels, exhibiting a much higher ignitability, with A-values ​​ranging from 0.1 to 10.0. By comparison, gasoline engines are ignitable with A-values ​​of 0.8 to 1.5, and gas engines up to an A-value of 3. Therefore, hydrogen mixtures often ignite during compression. In particular, even the smallest so-called hotspots on a spark plug or a hot oil particle can lead to ignition, so that fuel gas is already present in the cylinder before ignition. This results in high compression pressure, causing the rotation period to increase significantly with respect to a defined angular range. Furthermore, the intake pressure is very high in this case and exhibits high frequencies.Sometimes, a so-called diesel pilot, the injection of a defined quantity of diesel fuel, is used to ignite the hydrogen in the cylinder. This leads to controlled auto-ignition and can prevent premature ignition at a hotspot of the spark plug.

[0052] In an advantageous embodiment of the methods according to the first aspect or the second aspect, the state data each characterize a deviation of a first value of a parameter from a reference value of the parameter, wherein the parameters are the physical states or are derived from the physical states and wherein the reference value preferably occurs in an operation without combustion anomaly.

[0053] Preferably, the internal combustion engine is a reciprocating piston engine, and the first value of the parameter occurs during operation of a first cylinder of a reciprocating piston engine, and the reference value occurs during operation of other cylinders, in particular the remaining cylinders, of the reciprocating piston engine.

[0054] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, the rotation period is determined by means of a rotary encoder on a crankshaft of the internal combustion engine or the internal combustion engine to be analyzed, wherein measurement signals from missing grid elements of a scanning grid of the rotary encoder are supplemented in such a way that all grid elements describe the same angular sector. In one variant, measurement signals are supplemented such that they all occur after the same angular sector.

[0055] By supplementing the measurement signals of missing grid elements, a uniform sequence of measurement signals can be ensured.

[0056] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, an angular position of the missing grid elements is interpolated on the basis of a preceding grid element and a subsequent grid element.

[0057] By interpolating the missing grid elements, missing measurement signals can be added particularly easily.

[0058] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, the classifier is used to analyze pre-ignition, in particular early pre-ignition and / or late pre-ignition, as a combustion anomaly.

[0059] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, the operating data for generating the features further comprise a value profile of at least one measurement parameter from the following group of measurement parameters: a coolant temperature, an intake pressure, an intake temperature, an ignition angle with respect to the crankshaft position, an injection angle with respect to the crankshaft position, a knock sensor signal, a pressure in the exhaust manifold, an exhaust mass flow.

[0060] By taking into account further crankshaft angle-resolved or time-resolved measurement parameters, the accuracy of the classifier or the indirect measurement of the condition data can be increased.

[0061] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, in which the combustion anomalies to be detected are early pre-ignition and misfire, the operating data for generating the features further include a value profile of the intake pressure and preferably a value profile of the knock sensor signal.

[0062] Further consideration of the value profile of an intake pressure and the value profile of the signals from a knock sensor can differentiate early pre-ignition and misfire particularly well from other combustion anomalies.

[0063] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, in which the combustion anomalies to be detected are late pre-ignition, the operating data for generating the features further include a value profile of the knock sensor signal.

[0064] By considering the value profile of the knock sensor signal as a measurement parameter, late pre-ignition can be particularly well distinguished from other combustion anomalies.

[0065] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, an internal combustion engine is a reciprocating piston engine, in particular with spark ignition by means of a spark plug or diesel pilot ignition, wherein a data range of 30° to 90°, preferably about 60°, crankshaft angle for a six-cylinder engine, of 45° to 135°, preferably 90°, crankshaft angle for a four-cylinder engine, or of 60° to 180°, preferably about 120°, crankshaft angle for a three-cylinder engine is selected before and after the ignition top dead center (TDC) of a single cylinder in order to detect combustion anomalies of that cylinder.

[0066] By taking into account certain crankshaft angle ranges depending on the number of cylinders in a reciprocating engine, combustion anomalies of individual cylinders can be particularly well differentiated from other combustion anomalies.

[0067] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, an internal combustion engine is a reciprocating piston engine, in particular with spark ignition by means of a spark plug or diesel pilot ignition, wherein data ranges of 60° crankshaft angle for a six-cylinder, of 90° crankshaft angle for a four-cylinder or of 120° crankshaft angle for a three-cylinder are selected before and after the ignition top dead center (TDC) of all cylinders in order to detect combustion anomalies of all cylinders.

[0068] By excluding events from other cylinders, an unaffected evaluation of the signal can be performed. This allows for the analysis of a single cylinder or all cylinders of a reciprocating engine.

[0069] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, the data range for each cylinder is divided into two to six, preferably four, crankshaft angle sectors, wherein in each crankshaft angle sector a value of the change in rotation period is determined with respect to the respective crankshaft angle sector, wherein each value of the change in rotation period in conjunction with the respective crankshaft angle sector under consideration forms a feature.

[0070] By determining characteristics of the change in rotation period for specific crankshaft angle sectors, particularly meaningful characteristics can be generated.

[0071] In a further advantageous embodiment of the methods according to the first aspect or the second aspect, a slope is determined between a first value of the rotation period at the beginning of each crankshaft angle sector and a second value at the end of each crankshaft angle sector in order to calculate the change in the rotation period.

[0072] Calculating the slope using the given values ​​allows for a particularly simple determination of characteristics.

[0073] In a further advantageous embodiment of the methods according to the first aspect, the classifier is produced by a method according to the second aspect. The features and advantages described with respect to the first and / or second aspect of the invention also apply accordingly to the further aspects of the invention. Further features and advantages will become apparent from the following description of exemplary embodiments with reference to the figures. These show, at least partially schematically:

[0074] Figure 1: a functional block diagram of an embodiment of a system for detecting combustion anomalies in an internal combustion engine under analysis;

[0075] Figure 2: a flowchart of an embodiment of a method for detecting combustion anomalies in an internal combustion engine to be analyzed;

[0076] Figure 3: a functional block diagram of a system for generating a classifier for detecting combustion anomalies of an internal combustion engine;

[0077] Figure 4: a method for generating a classifier for detecting combustion anomalies of an internal combustion engine;

[0078] Figure 5: a detail of a rotary encoder for measuring a rotation period in relation to a defined angular range and a motor speed;

[0079] Figures 6a to 6d: Diagrams of cylinder pressure in conjunction with diagrams of rotation period for four angular ranges of the crankshaft, each with respect to top dead center; and

[0080] Figure 7: a supplemented diagram of a rotation period in relation to the ignition top dead center (TDC) from Fig. 6a to 6d.

[0081] Figure 1 shows a functional block diagram as an embodiment of a system 10 for detecting combustion anomalies in an internal combustion engine 1 under analysis. Figure 2 shows a flowchart of an embodiment of a method 100 for detecting combustion anomalies in the internal combustion engine 1 under analysis. Preferably, the method 100 for detecting combustion anomalies is carried out using the system 10 from Figure 1. Accordingly, the method 100 is explained below with reference to Figures 1 and 2. The following description refers to an application in which the internal combustion engine 1 under analysis is a hydrogen-powered reciprocating engine, hereinafter referred to as a hydrogen engine. However, it is easy for a person skilled in the art to transfer the described teaching to another type of internal combustion engine, in particular to another type of reciprocating engine.

[0082] In a first step 101 of the method 100 for detecting combustion anomalies, operating data of the hydrogen engine 1 to be analyzed, which is operated as a test specimen on a test bench or in a real vehicle 4, are recorded. Preferably, the value profiles of measurement parameters are measured via sensors 11. In this way, the operating data characterize the operating behavior of the hydrogen engine 1. At least one so-called rotary encoder is used as sensor 11, in which a rotation period is measured during which a drive shaft or crankshaft of the hydrogen engine rotates through a defined angular range, in particular 6° crank angle. In the box 11 of Figure 1, a section of a view of such a rotary encoder, also called an incremental encoder, is shown by way of example.

[0083] Furthermore, additional sensors 11 can be provided for acquiring measurement parameters. These additional sensors 11 include a knock sensor, a pressure sensor for measuring intake pressure, a temperature sensor for measuring coolant temperature and / or intake air temperature, a sensor for determining the ignition timing, and other sensors. In particular, the system 10 shown in Figure 1 can access engine control data, which is typically available via a bus system of an engine control unit.

[0084] In a second step 102, features are generated based on the recorded operational data, which are suitable as input data for a classifier 2. These features are generated, in particular, by processing at least a portion of the operational data using mathematical operations and / or by selecting a data range from the operational data. For this purpose, the system 10 preferably includes a data processing unit 12. Furthermore, the system 10 includes means 12a for generating the features.

[0085] Preferably, these are part of one of the data processing facilities 12.

[0086] In the box 12a shown in Figure 1, a slope with respect to a measurement curve is defined as a feature. Generating the features in this case preferably involves selecting the corresponding data range from the operating data and processing the operating data by means of a mathematical operation to calculate the slope or a derivative of the measurement curve with respect to the crankshaft angle (CW). Further signal processing for generating the features is possible. For example, signal processing can consist of supplementing measurement signals for missing grid elements of a scanning grid of a rotary encoder 11. Supplementing measurement signals ensures that all measurement signals occur after an identical angular sector. A detailed description of supplementing measurement signals is given below with reference to Figure 5.

[0087] Furthermore, when generating the features, the data ranges can be selected depending on the crankshaft angle relative to the ignition top dead center (TDC) of cylinders. For a six-cylinder engine, these data ranges are + / -60 degrees crankshaft angle (CW), for a four-cylinder engine + / -90 degrees crankshaft angle (CW) around the ignition top dead center, and for a three-cylinder engine + / - 120 degrees crankshaft angle before and after the ignition top dead center. Preferably, the combustion anomalies are analyzed individually for each cylinder.

[0088] The following features are determined: at least a change in the rotation period with respect to the defined angular range of the drive shaft of the hydrogen engine 1, a mean engine speed (i.e., the speed of the drive shaft or crankshaft of the hydrogen engine 1), and an engine torque of the hydrogen engine 1. The engine torque is preferably determined using an engine model, in particular a characteristic map, based on the recorded engine speed and control data of the hydrogen engine. In a third step 103, state data of the hydrogen engine 1 are finally determined by applying a classifier to the features generated in the second step 102.

[0089] The state data preferably characterize the physical states of the combustion chamber(s) of the hydrogen engine 1. The physical states describe the combustion. These physical states can preferably be pressure or heat, in particular thermal energy, present in the combustion chamber(s). The state data can therefore be corresponding values ​​of pressure and / or heat.

[0090] Further state data can be derived from, in particular, these measurement parameters. Examples of possible state data parameters are an indexing parameter of the cylinder pressure, which characterizes the internal torque of the internal combustion engine, specifically IMEP, i.e., an indicated mean effective pressure (IMEP) of the cylinders under consideration, and / or a so-called MFB50, i.e., a combustion center of gravity position or an angle relative to top dead center for the cylinders under consideration at which 50% of the fuel has been burned, a combustion peak pressure, or a heat release rate (RoHR). MFB50 is given in degrees of an angle, and IMEP is given in Pascals.

[0091] Combustion anomalies can be determined using such state data. Preferably, combustion anomalies of an operating combustion engine 1 under analysis are determined in relation to the operation of other combustion engines of the same type using reference values. In particular, the other combustion engines do not exhibit any combustion anomalies. More preferably, as an example, the operation of a first cylinder of a reciprocating engine under analysis is determined in relation to the combustion process of the other cylinders of the reciprocating engine under analysis. In particular, the other cylinders do not exhibit any combustion anomalies. In other words, a first cylinder 1, which exhibits a combustion anomaly, can be evaluated in relation to reference values ​​of state data from cylinders that do not exhibit a combustion anomaly.

[0092] Further state data can include the deviation of parameters, especially those mentioned above, from their respective reference values. The term "state data" will be used below for individual state data values.

[0093] For example, a condition datum Cyl1 Status. MFB50 can be determined by the deviation of the measured MFB50 in degrees of crank angle KW of cylinder 1 (MFB50_Cyl1) from the expected MFB50 (MFB50_Ref): MFB50_Ref is preferably an average MFB50 of all cylinders of a reciprocating engine that do not exhibit a combustion anomaly. An exemplary evaluation of the condition datum Cyl1 Status. MFB50 for this deviation could be as follows:

[0094] For example, the condition data Cyl1 Status. IMEP can be used for evaluation by calculating the relative percentage deviation of the measured IMEP of cylinder 1 (IMEP_Cyl1) from the expected IMEP (IMEP_Ref). IMEP_Ref can preferably be calculated as the average IMEP of all cylinders in a reciprocating engine that do not exhibit a combustion anomaly. An example evaluation of the condition data Cyl1 Status. IMEP for this deviation could be as follows:

[0095] Accordingly, further examples of state data include a relative deviation of RoHR (Rate of Heat Release) from the reference or a relative deviation of Cyl1 Status_PFP as a relative deviation of PFP (Peak Firing Pressure) from the reference. The assessments that characterize the physical states can also be considered state data within the meaning of the disclosure.

[0096] The listed parameters of a first cylinder 1 are preferably also determined for the remaining cylinders, so that a statement can be made about the quality of the combustion specifically for each cylinder and adjustments can be made directly to each cylinder individually, such as the adjustment of the injection quantity or the ignition angle.

[0097] Preferably, the determined status data are output in a fourth step 104, more preferably via an interface 14 of the system 10. The interface 14 can be a user interface or a data interface. Figure 1 shows, by way of example, the output of interface 14 of status data Cyl1 Status. MFB50, Cyl2 Status. MFB50, and Cyl3 Status. MFB50, as well as Cyl1 Status. IM EP, Cyl2 Status. lMEP, and Cyl3 Status. lMEP, where the dotted lines indicate that corresponding status data for other cylinders can also be output. Additional status data can also be output, which is further indicated by additional dotted lines. The determined or output status data can be used in a fifth step 105 to control a hydrogen engine 1. This allows the hydrogen engine 1 to be preferably protected or its function to be adjusted.

[0098] The following table shows the determination of control actions for a hydrogen engine 1 depending on the values ​​MFB50 and IMEP for a considered cylinder 1:

[0099] For example, Act_1 could mean that no action is performed, Act_2 that the load on cylinder 1 is reduced by reducing the fuel injection quantity for cylinder 1 by 10%, and Act_6 that the fuel injection quantity for cylinder 1 is completely deactivated.

[0100] Combustion abnormalities that can be detected using method 100 include pre-ignition, in particular early pre-ignition and / or late pre-ignition, knocking, delayed ignition or misfire.

[0101] Figure 3 shows a functional block diagram of a system 20 for generating a classifier 2 for detecting combustion anomalies in a hydrogen engine. Figure 4 shows a corresponding method 200 for generating a classifier 2 for detecting combustion anomalies in a hydrogen engine 1 to be analyzed. Preferably, the method 200 is carried out using the system 20. More preferably, the method 200 serves to generate a classifier 2, which can be used in the method 100 for detecting combustion anomalies in the hydrogen engine 2 to be analyzed. More preferably, the hydrogen engine 1 to be analyzed is of the same type as the hydrogen engines 1a, 1b, 1c, which are used to generate or train the classifier 2.The following section explains System 20 and Method 200 with reference to Figures 3 and 4, again purely by way of example with regard to the application case of hydrogen engines.

[0102] Method 200 is preferably carried out in a test bench operation of the hydrogen engines 1a, 1b, 1c. More preferably, the method can be carried out in a test operation with a suitably equipped test vehicle 4a, 4b, 4c. Accordingly, system 20 is part of a test bench or a test vehicle 4a, 4b, 4c, or is installed on a test bench or in a test vehicle 4a, 4b, 4c.

[0103] In the first step 201 of procedure 200, operating data from a plurality of hydrogen engines 1a, 1b, 1c of a specific type of hydrogen engine are recorded. This operating data includes the trend lines of measured parameters and characterizes the engine's behavior during operation. Furthermore, state data of each hydrogen engine are recorded during operation. Essentially, step 201 of procedure 200 is similar to the first step 101 of procedure 100. The difference, however, is that the state data of each hydrogen engine 1a, 1b, 1c are additionally recorded during operation.

[0104] In a second step 202, features for a classifier 2 are generated by processing at least a portion of the operational data using mathematical operations and / or by selecting a data range from the operational data. This step is essentially identical to the second step 201 of procedure 100.

[0105] In a third step 203, the classification algorithm 3 is trained using the generated features and their respective associated state data. This means that the values ​​of the state data that correlate with the generated features or the respective values ​​of the generated features are identified and / or provided to the classification algorithm 3. Preferably, the training of the classification algorithm 3 is carried out by repeatedly performing steps 201 to 203, each time acquiring the operating data and state data of a different internal combustion engine 1a, 1b, 1c.

[0106] As described above with reference to Figures 1 and 2, the state data used for training preferably characterize physical states of the combustion chamber(s) of the hydrogen engine 1 and are preferably determined separately for each cylinder. The physical states describe the combustion. Physical states can preferably be a pressure or a temperature, in particular a thermal energy, present in the combustion chamber(s). The state data can be corresponding values ​​of the pressure or temperature. Further state data can be derived from, in particular these, measurement parameters, as also described with reference to Figures 1 and 2. Examples of possible parameters of the state data are an indexing parameter of the cylinder pressure, which characterizes the internal torque of the combustion engine, in particular IMEP, i.e., an indicated mean effective pressure, e.g., IMEP, and / or a so-called MFB50, i.e.,The combustion center position or angle relative to top dead center, at which 50% of the fuel has been burned, a combustion peak pressure, or a heat release rate RoHR can be considered state data. Further state data can include the deviation of parameters, in particular the parameters mentioned above, from a respective reference value, as described with reference to Figures 1 and 2. The evaluations themselves can also constitute state data within the meaning of the disclosure. Preferably, the state data are determined and / or calculated during acquisition from the aforementioned measurement parameters and / or parameters, in particular from the physical states.

[0107] According to method 200, system 20, unlike system 10, has not only a rotary encoder 21 as a sensor, but also an additional sensor 22 for acquiring status data or physical states of the respective internal combustion engine 1a, 1b, 1c during operation. Such a sensor 22 can, in particular, be an internal cylinder pressure sensor. Furthermore, like system 10, system 20 has a data processing unit 23 with means 23a for generating features and means 23b for training the classification algorithm 3.

[0108] In method 200 for generating the classifier 2, the operating data and the state data are also generated by operating hydrogen engines 1a, 1b, 1c in a test bench environment as test specimens or in actual ferry operation in a vehicle 4a, 4b, 4c, as indicated in Figure 3. At the end of the training phase, the trained classification algorithm 3 is output as classifier 2, preferably via an interface 24, in a fourth step 204. Here, too, the interface 24 can be configured as a user interface or as a data interface. In a further embodiment, the rotary encoder 11 of system 10 and the rotary encoder 21 of system 20, as well as the data processing unit 12 of system 10 and the data processing unit 23 of system 20, are the same elements.

[0109] In principle, any type of rotary encoder can be used as rotary encoder 11 , 21 in the systems 10, 20, in particular an optical or electromagnetic rotary encoder.

[0110] In the exemplary embodiments, the classification algorithm 3 is an artificial neural network, and the classifier 2 is accordingly a trained artificial neural network. During training, weights and biases are adjusted so that the network is able to process the inputs correctly and generate the desired outputs. Preferably, the artificial neural network is trained using supervised learning. However, other training methods and / or classification algorithms 3 can also be used.

[0111] One possible first specific implementation is an artificial neural network trained to analyze combustion in all cylinders of the hydrogen engine 1 under analysis. Such an artificial neural network can be equipped with an input layer, five hidden layers, and an output layer. It outputs a diagnosis and information for detecting combustion anomalies in all cylinders. In this case, the input layer is fed with operating data and, during the training phase, with corresponding state data over two full crankshaft revolutions (DT1 to DT120 or 0° crank angle to 720° crank angle), where DT1 to DT60 correspond to the data of the first crankshaft revolution (DT1 to DT60) and DT61 to DT120 to the data of the second crankshaft revolution (DT1 to DT60).

[0112] A second specific embodiment involves the use of a set of simplified artificial neural networks, with a separate artificial neural network for each cylinder of the hydrogen engine. These simplified artificial neural networks preferably have an input layer, only two hidden layers, and an output layer. The output layer of each artificial neural network outputs only diagnostic information related to the detection of combustion anomalies in a single cylinder of the hydrogen engine. In other words, the output layer of each artificial neural network outputs only the data corresponding to a single cylinder, e.g., Cyl1.Statu. MFB50 for the neural network for cylinder 1. The input data consists solely of operating data and, if applicable, state data for that cylinder.Furthermore, to determine the cylinder's condition data, preferably only data from limited angular sectors before and after the ignition top dead center of the respective cylinder are considered, in particular + / -60. 0 , + / -90, or + / - 120°, depending on the number of cylinders.

[0113] Figure 5 shows a detail of an electromagnetic rotary encoder 11, 21 for measuring the rotation period with respect to a defined angular range and for measuring a motor speed. However, the principle for completing measurement signals explained with reference to Figure 5 can also be applied to other types of rotary encoders 11, 21.

[0114] In order to determine not only a relative but also an absolute position, rotary encoders 11, 21 generally have a reference mark. In the encoder wheel shown in Figure 5, this reference mark is a missing grid element of the scanning grid at position DT59. Accordingly, a measurement signal in the area of ​​the reference mark would correspond to three times the angular sector DT58* compared to normal measurement signals.

[0115] To compensate for the gap, the following procedure is preferably used in the example of a rotary encoder shown:

[0116] 1. DT1 , DT2, ... , DT57, DT58* correspond to the measured time values ​​between one edge of a grid element and the following edge, which correspond to an angular sector of 6° KW.

[0117] 2. Due to the lack of grid elements at positions DT59 and DT60, DT58* corresponds to a value of 18° KW and not 6° KW of an angular sector of the rotary encoder, as is the case with the other DT values ​​DT1 to DT57.

[0118] 3. The rotation duration values ​​DT58, DT59, and DT60 are calculated between the edges of the two modeled raster elements. The modeling is preferably performed as DT58 = DT59 = DT60 = DT58* / 3.

[0119] Figures 6a to 6d each show a diagram of the cylinder pressure value profile in conjunction with the same diagram of the rotation period value profile for four crankshaft angular ranges relative to a defined angular range as a function of the crankshaft angle from 60° crank angle before top dead center to 60° crank angle after top dead center. Using these combined diagrams, the following explains by way of example how features are generated from the measured rotation period relative to a defined angular range of the drive shaft, which indicate or characterize the changes in the rotation period relative to the defined angular range. The diagrams shown in Figures 6a to 6d are based on measurements of the cylinder pressure PZYL and the measurement of the rotation period for defined angular ranges of 6° crank angle, measured on a six-cylinder engine.

[0120] To generate the features, a data range is first selected from the operating data, corresponding to 60° crankshaft angle before and 60° crankshaft angle after the ignition top dead center of a single cylinder out of the 6 cylinders. This crankshaft angle range is particularly advantageous for a 4-stroke six-cylinder engine, as it roughly corresponds to the compression stroke and the power stroke of a cylinder.

[0121] To generate the individual features, this selected data range is further subdivided into four angular sectors, which correspond to smaller data ranges. Figure 6a shows a first data range for feature 1 of the cylinder under consideration, from a crankshaft angle KW1a of approximately 60° KW before ignition top dead center to a crankshaft angle KW1b of approximately 30° KW before ignition top dead center. Figure 6b shows a second data range for feature 2 of the cylinder under consideration, which extends from a crankshaft angle KW2a of approximately 30° KW before ignition top dead center to a crankshaft angle KW2b of approximately 10° KW before ignition top dead center. Figure 6c shows a data range for feature 3 of the cylinder under consideration, cyl1, which ranges from a crankshaft angle KW3a of about 10° KW after the ignition top dead center to a crankshaft angle KW3b of about 30° KW after the ignition top dead center.Figure 6d shows a data range for feature 4 with respect to the cylinder under consideration, cyl1, which extends from a crankshaft angle KW4a of approximately 30° KW after ignition top dead center to a crankshaft angle KW4b of approximately 60° KW after ignition top dead center. Based on the data ranges shown in Figures 6a to 6d, the change in the rotation period with respect to the defined angular range of 6° is calculated as the feature in each case. The rotation period with respect to a defined angular range DT(KW) is given in the lower diagram of Figures 6a to 6d.

[0122] To calculate the change in rotation period, the slope between the values ​​of the rotation period with respect to the defined angular range at the beginning of the selected data range for the crankshaft angle KW1 a, KW2a, KW3a, KW4a and the rotation period with respect to the defined angular range at the end of the selected data ranges for the respective crankshaft angle KW1 b, KW2b, KW3b, KW4b is preferably calculated.

[0123] With regard to feature 1 of the cylinder 1 under consideration, the diagram of the rotation period with respect to the defined angular range DT(KW) in Figure 7 is shown enlarged for illustrative purposes. According to this representation, the slope in the selected data range is given by the following equation:

[0124] (DT_lb - DT_l )

[0125] Feature 1st cylinder =

[0126] The first data range in Figure 6a corresponds to the beginning of the compression stroke. The second data range shown in Figure 6b corresponds to the second part of the compression stroke up to the ignition point. The third data range shown in Figure 6c corresponds to a period before and after reaching peak cylinder pressure. The fourth data range shown in Figure 6d corresponds to the period during the power stroke in which the pressure decreases. The rotation period DT plotted in Figure 7a corresponds to the defined angular range of 6° crankshaft angle.

[0127] It should be noted that the embodiments described are merely examples and are not intended to be restrictive in any way. Rather, the preceding description provides the skilled person with a guideline for implementing at least one embodiment, whereby various modifications, particularly with regard to the function and arrangement of the described components, can be made without departing from the scope of protection as defined by the claims and these equivalent combinations of features.

Claims

Patent claims 1. Computer-implemented method (100) for the indirect measurement of state data of an internal combustion engine (1) to be analyzed, in particular for the detection of combustion anomalies, by means of a classifier (2), comprising the following steps: • Acquisition (101 ) of operating data of the internal combustion engine (1) to be analyzed, wherein the operating data include value profiles of measurement parameters and characterize an operating behavior; • Generating (102) features for the classifier (2) by processing at least a part of the operational data by means of mathematical operations and / or by selecting a data range from the operational data; and • Determining (103) state data of the internal combustion engine (1) to be analyzed by applying the classifier (2) to the generated features; wherein the operating data comprise as measurement parameters a rotation period with respect to a defined angular range of a drive shaft and an engine speed of the internal combustion engine (1) to be analyzed, on the basis of which as features a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the internal combustion engine (1) to be analyzed are generated, and wherein the state data characterize physical states, in particular a pressure and / or a heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engine (1) to be analyzed.

2. Computer-implemented method (200) for generating a classifier (2) for determining state data of an internal combustion engine (1), in particular for detecting combustion anomalies of the internal combustion engine, by training a classification algorithm (3), comprising the following steps: • Acquisition (201 ) of operating data of a plurality of internal combustion engines (1 a, 1 b, 1 c) of a certain type of internal combustion engine, wherein the operating data comprise value profiles of measurement parameters and characterize behavior during operation, and of state data of the respective internal combustion engine (1 a, 1 b, 1 c) during operation; • Generating (202) features for a classifier (2) by processing at least a part of the operational data using mathematical operations and / or by selecting a data range from the operational data; and • Training (203) the classification algorithm (3) using the generated features and the respective associated state data, wherein the classifier (1) is generated; wherein the operating data comprise as measurement parameters a rotation period with respect to a defined angular range of a drive shaft and an engine speed of the internal combustion engines (1a, 1b, 1c), on the basis of which as features a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the respective internal combustion engine (1a, 1b, 1c) are generated, and wherein the state data characterize physical states, in particular a pressure and / or a heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engines (1a, 1b, 1c).

3. Method (100, 200) according to one of the preceding claims, wherein the internal combustion engine (1 ) is a hydrogen internal combustion engine.

4. Method (100, 200) according to one of the preceding claims, wherein the state data each characterize a deviation of a first value of a parameter from a reference value of the parameter, wherein the parameters are the physical states or are derived from the physical states and wherein the reference value preferably occurs in an operation without combustion anomaly.

5. Method (100, 200) according to claim 4, wherein internal combustion engine (1 ) is a reciprocating piston engine and the first value of the parameter occurs in an operation of a first cylinder of a reciprocating piston engine and the reference value occurs in an operation of other cylinders, in particular the remaining cylinders, of the reciprocating piston engine.

6. Method (100, 200) according to one of the preceding claims, wherein the Rotation period by means of a rotary encoder (11 ; 21 ) on a crankshaft of the internal combustion engines (1 a, 1 b, 1 c) or of the substance to be analyzed internal combustion engine (1 ) is determined, whereby measurement signals of missing grid elements of a scanning grid of the rotary encoder (11 ; 21 ) are supplemented in such a way that all measurement signals describe the same angular sector.

7. Method (100, 200) according to one of the preceding claims, wherein the Operating data for generating the features must also include a value profile of at least one measurement parameter from the following group of measurement parameters: a coolant temperature, an intake pressure, a Intake temperature, ignition angle in relation to crankshaft position, injection angle in relation to crankshaft position, knock sensor signal, exhaust manifold pressure, exhaust mass flow.

8. Method (100, 200) according to one of the preceding claims, wherein, for the analysis of early pre-ignition and misfire as combustion anomalies, the operating data for generating the features further comprise value profiles of the intake pressure and preferably of the knock sensor signal.

9. Method (100, 200) according to one of the preceding claims, wherein, for the analysis of late pre-ignition as a combustion anomaly, the operating data for generating the features further comprise a value profile of the knock sensor signal.

10. Method (100, 200) according to one of the preceding claims, wherein an internal combustion engine (1 , 1 a, 1 b, 1 c) is a reciprocating piston engine, in particular with External ignition by spark plug or diesel pilot ignition, wherein data ranges of 30° to 90°, preferably about 60°, crankshaft angle for a six-cylinder engine, of 45° to 135°, preferably 90°, crankshaft angle for a four-cylinder engine, or of 60° to 180°, preferably about 120°, crankshaft angle for a three-cylinder engine are selected before and after the ignition top dead center (TDC) of a single cylinder in order to detect combustion abnormalities of that cylinder.

11. Method (100, 200) according to one of the preceding claims, wherein an internal combustion engine (1 , 1 a, 1 b, 1 c) is a reciprocating piston engine, in particular with spark ignition by spark plug or diesel pilot ignition, wherein data ranges of 60° crankshaft angle for a six-cylinder, of 90° crankshaft angle for a four-cylinder or of 120° crankshaft angle for a three-cylinder are selected before and after the ignition top dead center (TDC) of all cylinders in order to detect combustion anomalies of all cylinders.

12. Method (100, 200) according to one of claims 7 or 8, wherein the data range for each cylinder is divided into two to six, preferably four, crankshaft angle sectors (KW1 , KW2, KW3, KW4), wherein in each crankshaft angle sector (KW1 , KW2, KW3, KW4) a value of the change in rotation period is determined with respect to the respective crankshaft angle sector, wherein each value of the change in rotation period in conjunction with the respective considered crankshaft angle sector (KW1 , KW2, KW3, KW4) forms a feature.

13. Method (100, 200) according to claim 9, wherein, for calculating the change in the rotation period, a slope is determined between a first value (KW1 a; KW2a; KW3a; KW4a) of the rotation period at the beginning of each crankshaft angle sector (KW1 , KW2, KW3, KW4) and a second value (KW1 b; KW2b; KW3b; KW4b) at the end of each crankshaft angle sector (KW1 , KW2, KW3, KW4).

14. Method (100, 200) according to claim 1, preferably in combination with one of claims 3 to 10, wherein the classifier 2 is produced by means of a method (200) according to claim 2.

15. System (10) for the indirect measurement of state data of an internal combustion engine (1) to be analyzed, comprising: • a rotary encoder (11 ) for recording operating data of the internal combustion engine (1) to be analyzed, wherein the operating data comprise a value profile of at least one rotation period with respect to a defined angular range of a drive shaft and a value profile of an engine speed of the internal combustion engine (1) to be analyzed as measurement parameters and characterize an operating behavior; • a data processing device (12) with means (12A) for generating features for a classifier (2) by processing at least a part of the operating data by means of mathematical operations and / or by selecting a data range from the operating data, wherein the features generated are a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the internal combustion engine (1) to be analyzed, and with means (12B) for determining state data of the internal combustion engine (1) to be analyzed by applying the classifier (2) to the generated features, wherein the state data characterize physical states, in particular a pressure and / or heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engine (1) to be analyzed.

16. System (20) for generating a classifier (2) for indirectly measuring state data of an internal combustion engine by training a classification algorithm (3) comprising: • a rotary encoder (21 ) for acquiring operating data of a plurality of internal combustion engines (1 a, 1 b, 1 c) of a type, wherein the Operating data include a value profile of a rotation period with respect to a defined angular range of a drive shaft and a value profile of an engine speed of the respective internal combustion engine (1 a, 1 b, 1c) as time-resolved measurement parameters and characterize an operating behavior; • a sensor (22) for recording state data of the respective internal combustion engine (1 a, 1 b, 1 c) during operation, wherein the state data characterize physical states, in particular a pressure and / or heat and / or an internal engine torque, of at least one combustion chamber of the internal combustion engines (1 a, 1 b, 1 c); and • a data processing device (23) with means (23A) for generating features for a classifier (2) by processing at least a part of the operating data by means of mathematical operations and / or by selecting a data range from the operating data, wherein the features generated are a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the internal combustion engines (1a, 1b, 1c), and with means (23B) for training the classification algorithm (3) using the generated features and the respective associated state data, whereby the classifier (2) is generated.

17. A computer-implemented classifier (2) for the indirect measurement of state data relating to combustion anomalies of an internal combustion engine, wherein the classifier (2) is generated by training a classification algorithm (3), wherein the classification algorithm (3) was configured by the following steps, which are performed for each training input of a plurality of training inputs: • Acquisition (201) of operating data of an internal combustion engine (1a, 1b, 1c) of a specific type of internal combustion engine, wherein the operating data comprise value profiles of measured parameters and characterize behavior during operation, and of Status data of the respective combustion engine during operation; • Generating (202) features for a classifier (1) by processing at least a part of the operational data using mathematical operations and / or by selecting a data range from the operational data; and • Training (203) the classification algorithm (3) using the generated features and the respective associated state data, whereby the classifier (1) is generated; wherein the operating data comprise as measurement parameters a rotation period with respect to a defined angular range of a drive shaft and an engine speed of the combustion chamber of the respective internal combustion engine (1a, 1b, 1c), on the basis of which a change in the rotation period with respect to the defined angular range, a mean engine speed and an engine torque of the respective internal combustion engine (1a, 1b, 1c) are generated as features, and wherein the state data comprise physical states, in particular a Characterize pressure and / or heat and / or internal engine torque, at least of one combustion chamber of the respective internal combustion engine (1 a, 1 b, 1 c).

Images