Diagnostics for monitoring structure-borne sound sensors in pressure profile analysis for emission modeling

Abstract

The present invention relates to a control unit for an internal combustion engine (20), configured to monitor an emission prediction in a cylinder of the internal combustion engine in real time, which is generated by means of a multi-stage emission model based on a sound signal from a structure-borne sound sensor, comprising: - Monitoring a stage within the multi-stage emission model to determine a pressure profile in the cylinder of the internal combustion engine based on the sound signal and comparison with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal when the corresponding monitoring criterion is not met. The invention further relates to a diagnostic method, a computer program, a diagnostic system, and a motor vehicle.

Description

[0001] The invention relates to a control unit, a diagnostic method, a computer program, a diagnostic system and a motor vehicle.

[0002] The invention lies in the field of engine technology, specifically control engineering for internal combustion engines. In the area of ​​monitoring and optimizing the combustion process, it is known to use signal data from a structure-borne sound sensor to model emission prediction in a cylinder of an internal combustion engine. Based on the signal data, the cylinder pressure profile can be determined, knocking can be detected, and appropriate action can be taken. These techniques are used to improve the efficiency and performance of internal combustion engines while simultaneously reducing emissions. With sufficient proof of the reliability of the emission prediction method, costly and difficult-to-provide sensors can be eliminated.

[0003] The reliability of existing acoustic emission prediction methods is not always sufficiently verifiable. Faulty or inaccurate acoustic emission predictions can lead to actual emissions exceeding permissible limits without this exceedance being recognized as unreliable. This can cause environmental pollution and lead to legal consequences if emission standards are not met. Therefore, there is a need for diagnostic and monitoring systems for acoustic emission prediction methods.

[0004] A well-known method to circumvent the need to install difficult-to-provide sensors is the evaluation of acoustic signals, which can be provided in many motor vehicles, for example, by a knock or structure-borne sound sensor.

[0005] For example, German patent application DE 10 2007 007 641 A1 describes a method for knock control in an internal combustion engine. A cylinder pressure characteristic is estimated from a cylinder pressure curve. This cylinder pressure characteristic is compared to a knock limit, which is a pressure value. The knock limit can be adapted via knock detection, which is based on a structure-borne sound signal. The structure-borne sound sensor is located on the internal combustion engine. The signal is then filtered using a bandpass filter. In addition to knock detection, the comparison can also be used to adapt the knock limit. Depending on the result, the knock limit is either maintained, increased, or decreased by a specific amount.

[0006] In publication US 2017 / 0051699 A1, a peak pressure value is determined using a knock sensor signal. A process is used to validate a predictive model for determining the peak pressure value. This process incorporates the predicted frequency bands, the predictive model, and the raw data signal from the knock sensor. First, the position of the peak pressure value is estimated using an algorithm. Then, the sensor signal is filtered with a low-pass or band-pass filter. Subsequently, the maximum, absolutely filtered signal values ​​for each frequency band are calculated, and the predictive model is applied to determine the predicted peak pressure values.

[0007] Document US 2017 / 0037798 A1 describes a system for controlling exhaust gas recirculation (EGR) flow in an internal combustion engine. This system uses a knock sensor. Based on a vibration signal from the knock sensor, the position and value of the peak pressure in the engine cylinder are determined. A diagnostic procedure is performed using an EGR knock flow rate modifier. This modifier defines a minimum and maximum expected value for the peak pressure. If the actual peak pressure value is outside the expected range, the modifier outputs a nominally adjusted target value to modify the combustion conditions within the cylinder.

[0008] For reliable diagnostics and emissions monitoring, it is advantageous to utilize the information content of an acoustic signal as comprehensively as possible. Furthermore, monitoring should be possible along the entire processing chain.

[0009] The object of the present invention is to provide a diagnostic method for monitoring an emission prediction method based on a multi-stage emission model based on a sound signal, which at least partially overcomes the aforementioned disadvantages.

[0010] This problem is solved by the control unit according to claim 1, the diagnostic method according to claim 13, the computer program according to claim 14, the diagnostic system according to claim 15 and the motor vehicle according to claim 16.

[0011] A first aspect of the invention relates to a control unit for an internal combustion engine, configured to monitor an emission prediction in a cylinder of the internal combustion engine in real time, which is generated by means of a multi-stage emission model based on a sound signal from a structure-borne sound sensor, comprising: - Monitoring a stage within the multi-stage emission model to determine a pressure profile in the cylinder of the internal combustion engine based on the sound signal and comparison with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal if the corresponding monitoring criterion is not met.

[0012] Further advantageous embodiments of the invention will become apparent from the dependent claims and the following description of preferred embodiments of the present invention.

[0013] Exemplary embodiments of the invention will now be described by way of example and with reference to the accompanying drawings. These show: Fig. 1 schematically an embodiment of a multi-stage method for generating an emission prediction from a sound signal; Fig. 2 schematically an embodiment of a first aspect of a diagnostic method according to the invention; Fig. 3 schematically an exemplary visualization of a transformed unfiltered frequency band and a filtered frequency band within the first aspect of the diagnostic procedure; Fig. 4 schematically an embodiment of a second aspect of the diagnostic method according to the invention; Fig. 5a and Fig. 5b schematically two exemplary visualizations of a pressure profile for an implementation of a tolerance range within the second aspect of the diagnostic procedure; Fig. 6 schematically an embodiment of a third aspect of the diagnostic method according to the invention; Fig. 7 schematically an embodiment for an application of the diagnostic aspects of the diagnostic method according to the invention in the multi-stage emission prediction method from Fig. 1; Fig. 8 a representation of an exemplary control unit which is designed to implement the diagnostic method according to the invention; and Fig. 9 a schematic representation of a motor vehicle according to the invention.

[0014] The preceding, more detailed description of embodiments of the present invention is followed by general remarks on the embodiments.

[0015] The exemplary embodiments show a control unit for an internal combustion engine. The control unit is designed to monitor an emission prediction in a cylinder of the internal combustion engine in real time. This emission prediction is generated by means of a multi-stage emission model that operates based on an acoustic signal from a structure-borne sound sensor. The control unit monitors one stage within the multi-stage emission model, whereby a pressure profile in the cylinder of the internal combustion engine is determined based on the acoustic signal. This profile is then compared with a monitoring criterion. If the corresponding monitoring criterion is not met, the generation of the emission prediction based on the monitored acoustic signal is prevented.

[0016] The control unit can be any device capable of monitoring emissions prediction. Preferably, but not necessarily, the control unit can operate in real time. For example, the control unit can be a vehicle control unit, in particular an electronic control unit (ECU) or an electronic control module (ECM). The control unit can also be part of another component. For example, the control unit is designed to receive and process sensor data from the structure-borne sound sensor in order to monitor and control various functions and components of the vehicle. This can also include the acquisition and processing of data from the structure-borne sound sensor as well as other relevant parameters of the internal combustion engine, such as cylinder pressure, engine speed, or temperature.

[0017] An internal combustion engine within the meaning of the present invention is a power engine that converts chemical energy into force or mechanical energy through combustion and is usually used to power a motor vehicle. In particular, an internal combustion engine is a piston engine, which is designed as a gasoline (spark-ignition) or diesel (compression-ignition) engine.

[0018] Monitoring can be performed in each cylinder of the internal combustion engine. Monitoring can be performed separately in each cylinder. The control unit can be used for an internal combustion engine with any number of cylinders. A structure-borne sound sensor can be provided for each cylinder. Alternatively, one structure-borne sound sensor can be provided for multiple cylinders.

[0019] Monitoring refers to monitoring during the operation of an internal combustion engine. The monitoring process steps require no external post-processing and are provided to the control unit during operation in such a way that the engine can be immediately restarted based on the monitoring data. Monitoring includes diagnostics for monitoring purposes. Within the monitoring process, a comparison with the monitoring criteria can be performed.

[0020] The pressure curve serves as the most significant parameter for inferring emission formation, regardless of the engine's age.

[0021] The monitoring criterion can be a condition that must be met before the sound signal can be used for further processing. The monitoring criterion can vary depending on the aspect.

[0022] An audio signal is a digitized representation of physical sound in a digitally processable form. Specifically, the audio signal is a sound signal in the frequency domain that has been transformed from a time-domain sound signal to the frequency domain using a Fourier transform within a sampling window (time interval).

[0023] The sampling window can be a predefined value, for example, 30 ms. Alternatively, the sampling window can be calculated based on the cadence or clock frequency of the control unit of the structure-borne sound sensor, the clock frequency of the internal combustion engine, or the frequency of the cylinder's duty cycle.

[0024] The transformation can be performed by the structure-borne sound sensor or the control unit. Accordingly, the sound signal can be captured by the control unit either as a time-domain signal or as a frequency-domain signal. The sound signal may be processed by filtering, signal decomposition, or component analysis.

[0025] The word "raise" refers to the recording, capturing or receiving of the sound signal from the structure-borne sound sensor by the control unit, whereby the sound signal is based on a measurement of the sound in the scanning window by the structure-borne sound sensor.

[0026] The word "determine" refers to the calculation, processing, or comparison of the sound signal from the structure-borne sound sensor by the control unit. Furthermore, the word "determine" refers to the calculation, processing, or comparison of the pressure profile determined based on the sound signal.

[0027] The structure-borne sound sensor, as disclosed, is a known sensor designed to detect vibrations, particularly high-frequency vibrations, in solids and convert them into electrical or digital signals. The structure-borne sound sensor is, for example, designed as a piezoelectric accelerometer and, as a broadband probe, covers a frequency range larger than that of typical knocking burns.

[0028] Emission prediction is a model that describes a range of exhaust gas products generated during the combustion process. These exhaust gas products include chemical compounds containing at least one of the following: nitrogen (N₂), carbon dioxide (CO₂), water (H₂O), carbon monoxide (CO), nitrogen oxides (NOx), ammonia (NH₃), methane (CH₄), formaldehyde (HCHO), nitrous oxide (N₂O), and hydrocarbons. The exhaust gas products may also include solid particles, such as soot.

[0029] Generating the emission forecast involves acquiring the sound signal and determining the pressure profile based on this signal, from which the emission forecast can be modeled. Additional input variables can be included in generating the emission forecast.

[0030] The multi-stage emission model is an emission model that generates emission predictions based on the sound signal from the structure-borne sound sensor through a sequence of several stages (steps). The emission model incorporates thermodynamic boundary conditions, some of which are not directly measurable or not directly measured, particularly the pressure profile, mass transfer, and temperature profile, derived from the pressure profile analysis.

[0031] In other words, emission prediction is a prediction of the exhaust gas products at the end of the combustion process in the internal combustion engine.

[0032] Emission prediction is based on real-time monitoring of a combustion process.

[0033] There are embodiments in which monitoring a stage includes monitoring the sound signal obtained from the structure-borne sound sensor, and the comparison with the monitoring criterion includes a comparison between a confidence value of the sound signal and an error threshold.

[0034] Monitoring the sound signal acquired from the structure-borne sound sensor can be a first step in a diagnostic monitoring process. The acquired sound signal can be a measured value output by the structure-borne sound sensor. The acquired sound signal from the structure-borne sound sensor can be referred to as the raw signal.

[0035] The confidence level indicates the degree of reliability or accuracy that the sound signal picked up by the structure-borne sound sensor can have.

[0036] The error threshold can represent a limit value that specifies the permissible deviation between the captured sound signal and an expected standard for the signal to still be considered reliable. The error threshold corresponds to the monitoring criterion, and the monitoring criterion is not met if the confidence level exceeds the error threshold. In a case where the confidence level exceeds the error threshold, the sound signal can be blocked from further processing.

[0037] By monitoring the sound signal measured by the structure-borne sound sensor and comparing it to an error threshold, the control unit can detect early on whether the detected sound signal is reliable and suitable for further processing. This improves the reliability of emission prediction and the efficiency of the internal combustion engine.

[0038] There are embodiments in which the acquired sound signal is transformed from the time domain into the frequency domain into an unfiltered frequency band.

[0039] An unfiltered frequency band is, for example, the entire parameter space of the frequency range.

[0040] Transforming the recorded sound signal into an unfiltered frequency band can be achieved using a Fourier transform, for example, a Fast Fourier Transform (FFT). This transformation converts the sound signal, which is originally in the time domain, into the frequency domain. This means that instead of considering the signal's amplitude as a function of time, the amplitude is now considered as a function of frequency.

[0041] Transforming the signal allows us to analyze the individual frequency components of the sound signal and determine which frequencies are dominant. The transformed sound signal, also known as a frequency spectrum, then contains information about the strength (amplitude) and phase of each frequency component in the sound signal.

[0042] The sound signal in the frequency domain can, according to the disclosure, be, for example, an energy spectrum or a power spectrum.

[0043] There are embodiments that further include defining relevant frequencies that are relevant for determining the cylinder pressure profile and filtering the unfiltered frequency band into a filtered frequency band using a bandpass filter arrangement based on the relevant frequencies.

[0044] The relevant frequencies (also referred to as "normal" frequencies) are those frequency ranges of the sound signal that are important for analyzing and determining the pressure profile in the internal combustion engine. These frequencies represent the typical oscillations and vibrations that occur during normal operation of the internal combustion engine. Any detected frequencies that are unknown under the given operating conditions of the internal combustion engine (described by speed, air consumption, and temperature) can limit the reliability of the sound signal and / or the structure-borne sound sensor.

[0045] Determining the relevant frequencies depends on various factors. Specific characteristics and operating conditions of the internal combustion engine can be considered. Furthermore, the relevant frequencies can be determined based on the engine order.

[0046] The bandpass filter arrangement can include one or more bandpass filters. A bandpass filter is a type of filter that suppresses or attenuates signal components whose frequencies lie outside one or more predetermined frequency ranges. In this case, the predetermined frequency range can correspond to the relevant frequencies. Aging or normal wear and tear (e.g., the break-in period of an internal combustion engine) can be taken into account when selecting the bandpass filter.

[0047] Bandpass filtering using a bandpass filter array allows the signal to be concentrated on the relevant frequencies. Furthermore, by filtering the signal into a filtered frequency band using a bandpass filter array, irrelevant frequencies can be identified, taking into account external influences (aging, wear).

[0048] There are embodiments that further include determining a noise component of the sound signal based on a comparison between the unfiltered frequency band and the filtered frequency band, and determining the confidence level of the sound signal based on the noise component.

[0049] The comparison between the unfiltered frequency band and the filtered frequency band can be made by calculating the difference.

[0050] The noise component is determined by comparing the unfiltered frequency band with the filtered frequency band. This comparison generates a difference spectrum. The difference spectrum represents the amount of noise or unexpected frequencies in the signal. This is the noise component.

[0051] For example, determining the noise component can involve comparing the energy of signal components in the filtered frequency band with the energy and / or power of signal components in the unfiltered frequency band.

[0052] If the signal in the frequency domain is an energy spectrum, then the energy of the signal in the filtered frequency band is calculated by integrating the energy spectrum over all filtered frequency bands. Similarly, the energy of the signal in the unfiltered frequency band is calculated by integrating the energy spectrum over all frequencies.

[0053] It should be noted that to calculate the energy of the signal in the unfiltered frequency band, an integral over the unfiltered components of all frequencies can also be used.

[0054] If the signal in the frequency domain is a power spectrum, the power is calculated mutatis mutandis in the filtered and unfiltered frequency bands.

[0055] The integral can be formed by applying a known numerical approximation.

[0056] The comparison between the energies and / or powers of the filtered and unfiltered signal components can be done, for example, by forming a quotient of the respective energies or powers.

[0057] The noise component indicates how much energy was removed by the bandpass filter arrangement. This component can be used to assess the quality and reliability of the audio signal. A low noise component suggests that the monitored audio signal consists primarily of the relevant frequencies. In this case, the monitored audio signal can be considered reliable. A high noise component indicates that the monitored audio signal contains many unexpected or interfering frequencies. In this case, the accuracy of emission predictions based on the signal may be compromised. Alternatively, the noise component can be determined by calculating the energy or power in the different frequency bands. The energy or power in a frequency band can be calculated by summing the squares of the amplitudes of the individual frequency components within that band.The noise component can then be defined as the ratio of the energy or power in non-relevant frequency bands (i.e., the frequency bands removed by filtering) to the total energy or power in the signal. These measures can reveal how much the amplitudes vary across the different bands and whether there are significant differences between the relevant and non-relevant bands. Alternatively, the noise component can be determined using signal processing or machine learning methods known to those skilled in the art. For example, pattern recognition or anomaly detection algorithms can be used to identify unexpected or abnormal frequency components in the signal.

[0058] The confidence level can be determined based on the noise component in various ways. The method for determining the confidence level should be chosen to account for specific requirements and circumstances. For example, the confidence level can be defined as the complement of the noise component, i.e., confidence level = 1 - noise component. Alternatively, the confidence level can be calculated based on a statistical analysis of the noise component. For instance, a distribution function (e.g., a normal distribution) could be fitted to the measured noise components, and the confidence level could be calculated as the percentile of the current noise component within this distribution. Another alternative approach is to calculate (determine) the confidence level by applying a suitable function to the noise component.

[0059] By monitoring the sound signal measured by the structure-borne sound sensor and comparing it to an error threshold, the control unit can detect early on whether the detected sound signal is reliable, suitable for further processing, and, if necessary, block it from further processing. This improves the reliability of emission prediction and the efficiency of the internal combustion engine.

[0060] There are embodiments in which monitoring a stage includes monitoring the plausibility of the pressure profile determined on the basis of the sound signal, and the comparison with the monitoring criterion includes a comparison between the determined pressure profile and a pressure profile calculated from a data-driven model, and a comparison of the comparison with a tolerance range.

[0061] Monitoring the plausibility of the pressure curve determined on the basis of the sound signal can be a second aspect of the diagnosis for monitoring the emission prediction.

[0062] A data-driven calculation can be implemented using a statistical model that maps a sound signal to a set of previously experimentally determined emission data. This data-driven model uses different data than the sound signal data from the structure-borne sound sensor.

[0063] The data-driven model can alternatively be implemented as a lookup table or as a semi-empirical model.

[0064] Plausibility refers to the credibility and reliability of the pressure profile determined based on the acoustic signal. If the monitoring criterion is met, the determined pressure profile is considered plausible and can be used for further processing. In this case, the determined pressure profile lies within the tolerance range.

[0065] The tolerance range can be known in advance. Each pressure value can be assigned a tolerance range. The tolerance range can vary along the pressure curve. Different input parameters can be included in the tolerance range. The tolerance range can be selected according to a specification of the internal combustion engine. Expected wear of the internal combustion engine can be taken into account when selecting the tolerance range.

[0066] In contrast to the first aspect of diagnostics for monitoring, the monitoring in the second aspect takes place at a different position within the processing chain of the multi-stage emission model. This allows for diagnostics to monitor the accuracy and reliability of the already processed sound signal from the structure-borne sound sensor.

[0067] In another embodiment, the first aspect described above can be used for monitoring, and the second aspect can be used for monitoring. In this case, two stages within the multi-stage emission model are monitored. Monitoring at one or both stages can, for example, be performed only on a sample basis. By combining both aspects, the reliability of the emission model based on the sound signal can be verified at two points within the processing chain of the multi-stage emission model. Furthermore, this allows a potential error within the multiple stages to be located.

[0068] There are embodiments that further include defining the tolerance range by including an upper error threshold and a lower error threshold, with plausibility monitoring carried out by comparing the result with the tolerance range.

[0069] The error thresholds can be known in advance. These thresholds represent limits that specify the permissible deviation between the two pressure curves for the determined pressure curve and the corresponding sound signal to still be considered plausible. The specific values ​​for the upper and lower thresholds can be defined based on the application's requirements. Various input variables can be used to determine the error thresholds. For example, at least one of the following can be included: a legal emission regulation, the engine's mileage, or a completed maintenance check of the engine. The input variables or error thresholds can be stored in memory.

[0070] Multiple process parameters can be compared over time and monitored using the tolerance range. These parameters can include, for example, the pressure curve in the cylinder, the engine speed, the air-fuel ratio, or the ignition timing.

[0071] By defining the tolerance range, including upper and lower error thresholds, the plausibility check can be specifically tailored. This allows for a more precise and reliable diagnosis.

[0072] There are embodiments in which the comparison of the determined pressure profile based on the sound signal with the calculated pressure profile based on the data-driven model takes place within a predefined time interval.

[0073] The time interval can encompass a period of time. Alternatively, the time interval can encompass a single point in time or multiple points in time. The time interval is assigned a tolerance range.

[0074] The time interval during a crankshaft revolution can begin with the intake valve closed and end with the ignition timing.

[0075] For example, using the known laws of thermodynamics, in particular those of polytropic processes, an expected target pressure at the ignition point can be determined based on this pressure. In compression-ignition engines, the point in time of the first discontinued injection can represent the end of the crankshaft interval under consideration.

[0076] In another embodiment, monitoring or comparison can only take place at one or more predefined points in time during ignition. In this case, the tolerance range can only be defined for the predefined point in time within the operating cycle under consideration.

[0077] By defining a predefined time interval, specific phases can be analyzed in a targeted manner. This enables precise and detailed monitoring and control of the measured pressure profile and thus the emissions. Furthermore, the predefined time interval ensures data comparability and optimizes computational effort.

[0078] There are embodiments in which determining the pressure curve based on the sound signal and calculating the cylinder pressure curve based on the data-driven model each involve different data.

[0079] The data-driven model for calculating the pressure curve can incorporate a wide variety of data collected during the operation of the internal combustion engine. This includes, for example, information about the engine's operating state, such as engine speed, injection timing, air-fuel ratio, ignition timing, and other relevant parameters. Similarly, historical operating data, environmental conditions, or specific engine characteristics can be integrated into the model to train the data-driven model.

[0080] This allows for an independent and additional perspective on the condition of the internal combustion engine. This increases the reliability and robustness of the diagnostic monitoring system, as faults or deviations in the acoustic signals from the structure-borne sound sensor can be detected and compensated for by the independent data source.

[0081] There are embodiments in which monitoring a stage includes monitoring the generated emission prediction, and the comparison with the monitoring criterion includes a comparison between the generated emission prediction and a measured value from an emission sensor and a comparison of the comparison with a confidence interval.

[0082] Monitoring the generated emission forecast can be a third aspect of the diagnostic process for monitoring the emission forecast.

[0083] The emissions sensor could be, for example, a NOx sensor. It could also include one or more of the following sensors: lambda sensor, temperature sensor, pressure sensor, mass airflow sensor, particulate sensor, CO2 sensor, or fuel pressure sensor. The measured values ​​can be acquired at specific points or continuously. The calibration process can begin after a predefined period, such as after the combustion engine has been started.

[0084] The monitoring criterion can include monitoring limits. These limits can be defined based on a previously known measurement uncertainty of the respective sensor and a previously known uncertainty of the multi-stage emission model. Coupled from this, the confidence interval can be defined. The confidence level can represent the smallest possible error threshold for monitoring the emission prediction.

[0085] The sensor's measurement uncertainty may be known, for example, from a technical specification provided by the manufacturer. The model uncertainty can be determined, for example, during parameterization.

[0086] The third aspect of monitoring can serve as an optional addition to the first and / or second aspect of the diagnostic procedure.

[0087] This allows the actual pollutant concentration generated (by the emission forecast) to be compared with a measured pollutant concentration. This enables the reliable verification of the emission forecast's functionality using real-world emission data. Furthermore, the use and comparison with other sensor data allows for additional independent monitoring of the emission forecast.

[0088] There are embodiments in which the comparison between the generated emission prediction and the measured value of the emission sensor takes place in a transition range to an operational readiness of the emission sensor.

[0089] The transition range refers to the phase in which the emission sensor moves from an inactive or cold state to a fully functional state. During this transition range, the sensor warms up and begins to provide reliable measurement data.

[0090] The third aspect of the diagnostic procedure allows for targeted monitoring of the transition period until the emission sensor is fully operational and the pollutant emissions from the combustion engine stabilize after a start-up phase. This enables the reliability of the multi-stage emission prediction method based on the sound signal to be demonstrated even in this transition period.

[0091] There are embodiments that also include issuing an error message if the monitoring criterion is not met.

[0092] The error message could be a customer restriction. This customer restriction could be legally regulated. For example, the customer restriction could be analogous to an SCR system, which is common in modern diesel engines. The error message could also be a warning message displayed to the driver.

[0093] The error message could be a fault code entry. This fault code entry could, for example, include an OBD-2 fault code.

[0094] This ensures that only reliable sound signals from the structure-borne sound sensor are used to generate the emission prediction.

[0095] A second aspect of the invention relates to a diagnostic method for monitoring an emission prediction in a cylinder of an internal combustion engine in real time, which is generated by means of a multi-stage emission model based on a sound signal from a structure-borne sound sensor, comprising: - Monitoring a stage for determining a cylinder pressure profile in the cylinder of the internal combustion engine based on the sound signal within the multi-stage emission model and comparison with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal if the corresponding monitoring criterion is not met.

[0096] The method according to the invention can be carried out by a control unit of a conventional motor vehicle, provided that the control unit has the necessary computing capacity.

[0097] A third aspect of the invention relates to a computer program for monitoring an emission prediction in a cylinder of an internal combustion engine in real time, which, when executed by a computer, causes the computer to perform the following method: - Monitoring a stage for determining a cylinder pressure profile in the cylinder of the internal combustion engine based on the sound signal within the multi-stage emission model and comparison with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal if the corresponding monitoring criterion is not met.

[0098] A computer program can be digitally integrated into a motor vehicle's control system, so that the control system is configured to execute the method according to the invention.

[0099] A fourth aspect of the invention relates to a diagnostic system for an exhaust gas monitoring system for real-time monitoring of a combustion process, for monitoring the exhaust gas monitoring system, comprising the control unit according to the invention and the structure-borne sound sensor.

[0100] The combustion process refers to the combustion process resulting from ignition during a working cycle of a cylinder in an internal combustion engine. The combustion process can also be referred to as the working cycle or power cycle.

[0101] A fifth aspect of the invention relates to a motor vehicle according to the invention, comprising - the internal combustion engine, the exhaust gas monitoring system and a diagnostic system according to the invention.

[0102] The motor vehicle can be a passenger car. The motor vehicle can be a truck. The motor vehicle can be a motorcycle. The motor vehicle can be a bus. The motor vehicle can be a watercraft.

[0103] Emission prediction through analysis of sound signals in internal combustion engines: To increase the efficiency of internal combustion engines while minimizing environmental impact, the multi-stage process described below enables comprehensive, real-time emission prediction through the precise analysis of sound signals acquired by a structure-borne sound sensor on the internal combustion engine. By converting sound signals into detailed pressure profiles in the combustion chamber and processing this data using advanced methods such as machine learning and physical models, the process provides a basis for predicting and controlling emissions from internal combustion engines.

[0104] Returning to the attached figures, it shows Fig. Figure 1 schematically illustrates an embodiment of a multi-stage method for generating an emission prediction from an acoustic signal. The emission prediction serves for real-time monitoring of exhaust gas components from a combustion process in a cylinder of an internal combustion engine 20 of a motor vehicle 100. In this method, a pressure profile (cylinder pressure profile) of the internal combustion engine 20 is determined using a structure-borne sound sensor 10. An acoustic signal is detected by the structure-borne sound sensor 10, which is attached to the internal combustion engine 20. From this acoustic signal, a pressure profile in the combustion chamber of the internal combustion engine 20 is derived using specific processing methods, such as machine learning (ML) and physical models. This pressure profile is analyzed to generate an emission prediction that describes the composition of the exhaust gas components.

[0105] The process for generating the emission prediction takes place in several stages, which are based on the evaluation of the pressure profile in the combustion chamber of the internal combustion engine 20, derived from the sound signal. The individual stages ST1 to ST4 are described below. Stages ST1 to ST4 are in Fig. 1 each represented as a block one below the other and connected to each other by an arrow.

[0106] In investigation stage ST1, a raw signal (sound signal) is acquired or measured by the structure-borne sound sensor 10 during operation, such that the raw signal represents at least one combustion process of the internal combustion engine 20. The structure-borne sound sensor 10 is located in the vicinity of the internal combustion engine 20. For example, the structure-borne sound sensor 10 can be attached to the crankcase of the internal combustion engine 20, so that sound generated during the combustion process can be measured by the structure-borne sound sensor 10. The structure-borne sound sensor 10 detects the sound waves generated during the operation of the internal combustion engine 20, which are caused by vibrations and oscillations within the engine 20. These sound waves are converted by the structure-borne sound sensor 10 into a sound signal, which can then be further processed.

[0107] In transformation stage ST2, the raw signal is transformed and modeled. The raw signal from stage ST1 of the structure-borne sound sensor 10 is provided either in the time domain or the frequency domain. If the raw signal is a sound signal in the time domain, it is transformed into a sound signal in the frequency domain. For this purpose, a sampling window is defined. The sampling window is a time window of a measured sound signal that is to be transformed into the frequency domain. The sampling window can be, for example, 30 ms, but longer or shorter sampling windows are also possible according to the invention. For example, the sampling window can encompass the duration of a work cycle. The transformation into the frequency domain can be performed using a Fast Fourier Transform (FFT). The raw signal transformed into the frequency domain can be referred to as the frequency raw signal.

[0108] The ST2 transformation stage also includes signal conditioning. This signal conditioning process the raw frequency signal (e.g., through filtering, signal decomposition, component analysis, etc.) to isolate relevant frequency ranges. The signal conditioning identifies and isolates the signal components relevant for reconstructing the cylinder pressure profile, thereby increasing model accuracy.

[0109] In stage ST3 of the analysis, a pressure profile in the combustion chamber of the internal combustion engine 20 is derived from the processed sound signal. This profile represents the pressure within the cylinder over the course of a combustion cycle. The sound signal is first isolated to relevant frequency ranges and then fed into the corresponding model to determine the pressure profile. Determining the pressure profile from the sound signal can be done in various ways. A physical model can be used that maps the sound signal to the pressure profile in the cylinder. This model can, for example, be based on a mathematical model that describes the transmission behavior between sound and pressure.

[0110] Alternatively, a machine learning model can be used that maps the sound signal to a pattern pressure curve. This model is first trained with training data. The training data contains experimentally determined pressure curves with associated sound signals. Again, the model can be a pure machine learning model that determines the pressure curve from the sound signal based on training data, without mapping the sound signal to a pattern pressure curve.

[0111] The model infers the cylinder pressure from the sound signal, which was acquired and, if necessary, processed in the previous stages ST1 and ST2. The cylinder pressure as a function of time is a pressure curve. This pressure curve is determined based on the sound signal.

[0112] In emissions calculation stage ST4, the pressure profile determined in stage ST3 is analyzed. Stage ST4 then yields an emissions prediction, which is available for further use. The emissions calculation in stage ST4 can be performed in different ways. For example, the pressure profile determined in stage ST3 can first be used to infer the temperature profile and the mass conversion of the fuel in the combustion chamber. Based on the temperature profile and the mass conversion, the reaction kinetics of the relevant chemical reactions in the combustion chamber are calculated numerically, thus determining the remaining emissions at the end of a combustion process. Alternatively, the numerical calculation of the reaction kinetics can be replaced by a machine learning (ML) model. The numerical calculation of the reaction kinetics then provides the relevant training data for the ML model.Alternatively, characteristic parameters are calculated from the determined pressure profile from stage ST3 (e.g., maximum cylinder pressure, maximum cylinder pressure gradient, etc.) and correlated with measured emission values. Using a machine learning model and a training database, emission formation is thus calculated directly from a modeled cylinder pressure profile. The thermodynamic evaluation of the pressure profile can be omitted. The models predict the concentrations of various exhaust gas components such as CO2, NOx, CO, and others, based on the condition of the internal combustion engine 20 and the behavior of the combustion process. The emission prediction can be used to optimize the combustion process in real time.By adjusting parameters of the internal combustion engine 20 such as injection timing, air-fuel ratio and ignition timing, the operation of the internal combustion engine 20 can be controlled in such a way as to minimize emissions and optimize performance. Diagnostic method for optimizing internal combustion engines through frequency analysis of the structure-borne sound sensor - First aspect

[0113] The following describes a first aspect of a diagnostic method according to the invention, which relates to the monitoring of the raw signal acquired by the structure-borne sound sensor 10 for the optimization of internal combustion engines.

[0114] Such a diagnostic procedure can contribute to improving the performance and efficiency of internal combustion engines, reducing emissions, extending the engine's service life, and saving costs. Furthermore, the diagnostic procedure can improve the reliability of the method for generating emission predictions from a sound signal.

[0115] In the first aspect of the diagnostic procedure, the raw electrical signal from the structure-borne sound sensor 10 is transformed into the frequency domain, and a set of relevant frequencies for generating the pressure curve is determined. For the given operating condition of the internal combustion engine, described by factors such as engine speed, air consumption, and temperature, all unknown frequencies are identified. These unknown frequencies limit the reliability of the raw signal from the structure-borne sound sensor.

[0116] The method focuses on examining the first motor order and its multiples via bandpass filters. By calculating the difference between the unfiltered and filtered frequency bands, an acceptable level of interference can be verified.

[0117] The higher the proportion of existing interference frequencies, the lower the confidence level. By comparing the determined confidence level with an error threshold, a decision can be made as to whether the signal is further processed for determining the pressure profile. Comparing the confidence level with the error threshold is equivalent to comparing it with a monitoring criterion.

[0118] Fig. Figure 2 shows an exemplary first aspect of the diagnostic method according to the invention for the multi-stage method for generating an emission prediction from a sound signal according to Fig. 1. The first aspect of the diagnostic procedure can be, for example, according to stage ST1, while it is stage ST2 or after stage ST2 of the procedure according to Fig. 1. Generally, the diagnostic procedure is carried out during operation of the motor vehicle 100. In the procedure according to Fig. 2. The raw signal acquired by a structure-borne sound sensor 10 is continuously recorded and analyzed. The raw signal continuously provides basic information by means of which conclusions about the condition, emission and performance of the internal combustion engine 20 can be drawn during its operation.

[0119] In a first step S21 of the first aspect of the diagnostic procedure, the raw electrical signal from the structure-borne sound sensor 10 is transformed into the frequency domain (frequency band). This transformation is carried out, for example, by applying a Fourier transform, preferably a Fast Fourier Transform (FFT), as is known to those skilled in the art. After the transformation, the transformed raw signal can be analyzed by examining the amplitude and phase of the individual frequency components.

[0120] In a subsequent step S22, relevant frequencies are determined for a given operating state of the internal combustion engine. The selection of the relevant frequencies (e.g., frequency bands) is based on knowledge of the normal operating conditions of the internal combustion engine. The relevant frequencies differ depending on the design and type of the internal combustion engine. Frequencies that typically occur during normal operation of the internal combustion engine are considered relevant. These represent the usual vibrations and oscillations that occur during operation. In summary, in step S22, the frequencies relevant for analysis are selected within the frequency domain of the transformed raw signal. In this case, the selected ranges ("bands") are chosen to cover the first engine order and its multiples.The relevant frequencies can be specifically known through the selection of the internal combustion engine. These frequencies can also be stored in a memory for selection. In a subsequent step S23, the transformed raw signal is examined for disturbances or noise (noise component) that could impair the accuracy of a later emission prediction. A check for an acceptable noise component can be performed by calculating the difference or quotient between the unfiltered and filtered frequency bands. This determines the amount of noise or unexpected frequencies in the transformed raw signal. The unfiltered raw signal is the transformed raw signal as obtained in step S21. It contains all frequency components detected by the structure-borne sound sensor 10, including any disturbances or noise.A filtered frequency band is derived from the unfiltered frequency band using the bands determined in step S22. The filtered frequency band contains only the frequency components within the (pre-defined) frequency bands around the fundamental frequency and its multiples. By calculating the difference between these two signals (the unfiltered frequency band and the filtered frequency band), a difference spectrum is determined, representing the amount of noise or unexpected frequencies in the signal. The difference spectrum represents the noise component in the raw signal.

[0121] In a subsequent step S24, a confidence level is determined based on the noise component determined in S23. The higher the proportion of existing noise frequencies, the lower the confidence level of the structure-borne sound sensor 10. This value indicates the degree to which the captured raw signal is reliable for further processing.

[0122] The confidence level indicates the degree of reliability or accuracy that the raw signal acquired and transformed by the structure-borne sound sensor 10 can have. If the system detects frequencies that are unknown or unexpected, the confidence level is reduced. This can occur, for example, if the detected frequencies (described by engine speed, air consumption, and temperature) do not correspond to the normal operating conditions of the internal combustion engine, which were previously defined in S22. This means that the system has less confidence in the accuracy or reliability of the (transformed) raw signal. Any detected frequencies that are unusual in the current operating state of the internal combustion engine (described by factors such as engine speed, air consumption, and temperature) reduce the confidence level.This means that frequencies outside the relevant range previously defined in S22 are considered potential indicators of interference or problems.

[0123] In one embodiment where the noise components are described as a difference spectrum, the confidence level can be derived directly from the difference spectrum. The difference spectrum is analyzed to calculate a quantitative value for the noise component. For example, the sum or integral of the amplitudes in the difference spectrum is calculated. This yields a value representing the overall strength of the noise frequencies in the signal. If this value is small, it means that the unfiltered and filtered signals are very similar and that the signal contains little noise. In this case, the confidence level is high. However, if the value is large, it means that there are many noise frequencies in the signal, and consequently, the confidence level is low.

[0124] Alternatively, the confidence level can be calculated by determining the proportion of interference frequencies to the total energy of the signal. This can be achieved by squaring and summing the amplitudes in the difference spectrum and dividing by the sum of the amplitudes in the original signal.

[0125] In a subsequent step S25, the confidence level is compared to an error threshold. This corresponds to a comparison with a monitoring criterion. The determined confidence level is compared to a predefined error threshold. If the confidence level exceeds this threshold, this may indicate a problem with the raw signal, the structure-borne sound sensor 10, or the internal combustion engine 20. In this case, the system can decide to stop (lock) further processing of the raw signal or at least issue a warning that the data may be unreliable. This serves to exclude unreliable or erroneous data from further analysis for emission prediction. This step helps to monitor data quality and ensure that the analysis is based on reliable and accurate data.

[0126] By monitoring and analyzing the raw signal from the structure-borne sound sensor 10, as described in the first aspect of the diagnostic procedure above, malfunctions and anomalies can be detected early, and the engine operation can be adjusted accordingly. This can help improve the efficiency and performance of the internal combustion engine. The diagnostic procedure also enables accurate and reliable emission prediction, which can be used for real-time optimization of the combustion process. This can help minimize harmful emissions. Continuous monitoring of the structure-borne sound sensor 10 allows potential problems or defects to be detected and rectified early, before they lead to major damage or breakdowns. Furthermore, early detection and resolution of problems can help avoid costly repairs or downtime.Furthermore, the process can help reduce fuel consumption, leading to cost savings. By optimizing the operation of the internal combustion engine and detecting problems early, the engine's service life can be extended.

[0127] In the exemplary embodiment of the Fig. 2. Based on the noise components, a confidence level is determined, indicating the degree to which the captured raw signal is reliable for further processing. In other applications, it may be advantageous to choose a more complex representation of the confidence level. Instead of using a single value, other metrics or a multidimensional representation of confidence can be used to gain more detailed insights into the quality and reliability of the signal.

[0128] Instead of a single confidence value, a confidence interval can be specified, indicating a range within which the true value lies with a certain probability. This can better account for uncertainties in measurement and analysis. For more complex systems, a matrix can be used to represent different aspects of confidence. For example, different rows of the matrix can represent different frequency bands, and the columns can represent different statistical measures such as the mean, variance, or skewness of the disturbance frequencies in each band. A confidence profile can be a graphical representation showing how confidence changes across different parameters or conditions. For example, a profile can show how confidence changes with engine load or engine speed.In dynamic systems, confidence can be modeled as a function of time, with the confidence value updated at each point in time based on current and past signal values. In systems using multiple sensors, a combined confidence value can be calculated from the individual sensor values, possibly using methods such as fuzzy logic or machine learning to account for the interactions between the sensors. These more complex representations of trust can enable precise and nuanced control and monitoring of systems, especially in critical or variable environments. Consideration of the motor order to determine the usual frequencies

[0129] The following is an example of how to determine the relevant ("usual") frequencies (step S22 in Fig. 2) described. According to this embodiment, the engine order is considered. The first engine order and its multiples are analyzed using bandpass filters. This allows for a targeted analysis of frequency ranges that may be relevant for engine operation. In this context, the engine order refers to the fundamental vibrations or frequencies generated by the normal operation of the internal combustion engine. The first engine order corresponds to the fundamental frequency generated by a complete engine cycle. Multiples of this fundamental frequency represent higher engine orders.

[0130] In step S22, the first engine order and its multiples are specifically examined. This is achieved by applying bandpass filters that only allow frequencies within a specific range to pass through. This means that the system specifically analyzes the vibrations or frequencies generated by the normal operation of the internal combustion engine. Other frequencies that do not belong to these engine orders are filtered out. This targeted analysis of the frequency ranges relevant to engine operation can help improve the accuracy and reliability of the data analysis. By analyzing only the relevant frequencies, the system can make a more accurate prediction of engine behavior and pressure curves.

[0131] The fundamental frequency of an internal combustion engine, also known as the first order, is closely related to the engine speed. It describes the number of complete power cycles the engine performs per unit of time. In a four-cylinder, four-stroke engine, for example, where each cylinder completes one power stroke per revolution of the crankshaft, the fundamental frequency corresponds to half the engine speed. If the engine is running at 3,000 revolutions per minute (rpm), for instance, the fundamental frequency is 1,500 cycles per minute or 25 cycles per second (Hertz).

[0132] The rotational speed of the internal combustion engine can be detected, for example, by a speed sensor. This sensor is usually connected to the vehicle's engine control unit (ECU), which uses the speed information for various purposes, such as engine control and monitoring.

[0133] The fundamental frequency f (also called the first engine order) of the engine (internal combustion engine) can be calculated from the engine speed N. The specific formula depends on the type of engine. For a four-cylinder four-stroke engine, in which each cylinder performs one working stroke every two revolutions of the crankshaft, the formula would be: f=N / 2*˙(1 / 60)

[0134] Here, f is the fundamental frequency in Hertz (Hz) and N is the rotational speed in revolutions per minute (rpm).

[0135] The formulas above assume an ideal situation. In other implementations, further factors may be considered, such as imbalances, mileage, mechanical aspects, and other factors that influence the actual fundamental frequency.

[0136] The multiples of the fundamental frequency, also called overtones or harmonics, can be easily determined by multiplying the fundamental frequency by integers. Assume the fundamental frequency of an engine (the first engine order) is f. The second engine order would then be 2f, the third 3f, the fourth 4f, and so on. For example, if the fundamental frequency is 25 Hz (which would be the case for a four-cylinder, four-stroke engine running at 3000 rpm), then the second engine order would be 2 * 25 Hz = 50 Hz, the third 3 * 25 Hz = 75 Hz, the fourth 4 * 25 Hz = 100 Hz, and so on. These multiples of the fundamental frequency represent higher-frequency components of the engine signal, which can be generated by various mechanical and acoustic phenomena within the engine. They can often provide important information about the engine's condition and behavior.

[0137] Frequency bands can be created from the first-order motor frequencies and their multiples by defining ranges around each of these frequencies. Each band focuses on a specific frequency (e.g., the fundamental frequency or its multiples) and a specific range around that frequency. Using such frequency bands allows the signal to be analyzed in specific frequency ranges relevant to motor operation. This can help improve the accuracy and reliability of the analysis, especially when the signal has complex or multiple frequency components.

[0138] Fig. Figure 3 schematically shows an exemplary visualization 50 of a transformed unfiltered frequency band 52 and a filtered frequency band 53 of the first aspect of the diagnostic procedure. Based on Fig. 3. The determination of the interference component is visually represented, which is essentially part of the first aspect of the previously described diagnostic procedure. Fig. 2 corresponds. Fig. Figure 3 shows an example of an unfiltered signal power spectrum of the structure-borne sound sensor 10 in the frequency domain, and a corresponding filtered signal power spectrum. Frequency bands 52 and 53 each represent an amplitude value (signal strength) (axis A) versus a frequency (axis f) in a separate diagram. The frequency axis f can be subdivided into different sections according to the engine order of the internal combustion engine. For example, in Fig. 3. The subdivision according to the engine order into 1f, 2f and 3f. The subdivision is as follows: Fig. 3 is represented by vertical dashed lines. In this case, the first motor order is considered as an example. The first motor order 1f is represented by section 51 along the frequency axis f. The unfiltered frequency band 52 is shown in the upper diagram. The unfiltered frequency band 52 serves as the output value in step S21 in Fig. 2, in which it is determined from the sound signal of the structure-borne sound sensor 10 via a transformation. The filtered frequency band 53 is shown in the lower diagram. The filtered frequency band 53 serves as the output value in step S23 in Fig. 2, in which it is determined by bandpass filtering using the bands determined in step S22. To determine the noise component, the energy contents of the unfiltered frequency band 52 and the filtered frequency band 53 are then determined, for example, by integration within the first motor order 1f. The resulting integrals are then compared. This allows the noise component to be quantified, which is output as a value in step S23. Fig. 2. The noise component contains information about how much energy was removed by the bandpass filter. Based on the noise component, the monitored sound signal can be assigned a confidence value. The confidence value serves as the output value in step S24 in Fig. 2. Monitoring the plausibility of the determined cylinder pressure curve - Second aspect

[0139] The second aspect of the diagnostic procedure, described below, enables monitoring of the cylinder pressure profile (pressure curve) to optimize engine performance and reduce emissions. In this procedure, the pressure profile of the internal combustion engine 20, determined based on the raw signal, is continuously monitored and compared with a predicted (calculated), data-driven model. This model uses historical and current operating data, not derived from the structure-borne sound sensor 10, to calculate an expected pressure profile. By comparing the measured values ​​obtained from the multi-stage procedure with the calculated model predictions within a defined tolerance range, the plausibility of the determined pressure profile can be monitored and assessed. This comparison corresponds to a comparison with a monitoring criterion.Deviations that fall outside these limits signal potential problems that may require further investigation or adjustments.

[0140] Fig. Figure 4 shows an example flowchart for the second aspect of the diagnostic procedure.

[0141] In step S31, the pressure profile is determined. The pressure profile in the internal combustion engine 20 during one crankshaft revolution is calculated according to stage ST3 of the previously described multi-stage procedure. Fig. 1. An emission prediction is determined from a sound signal. This occurs in the period from the moment the intake valve closes ("intake valve closed") until the moment of ignition ("ignition time").

[0142] In the subsequent step S32, an expected pressure profile is calculated using data-driven modeling. A data-driven model is used to calculate the expected pressure profile in the internal combustion engine 20. This is based on data collected during the operation of the internal combustion engine 20. The input data for the data-driven model to calculate the pressure profile are different from those used to determine the pressure profile in step S31, specifically not the signal data from the structure-borne sound sensor 10.

[0143] In the subsequent step S33, the determined pressure profile is compared with the calculated pressure profile. The determined pressure profile is compared with the calculated pressure profile, which was generated by the data-driven modeling. Within a predefined time interval—namely, the period from the intake valve closing to the ignition point—each value of the data-driven (calculated) pressure profile is compared with the corresponding value of the pressure profile determined by structure-borne sound sensor 10. If the deviation between the two pressure values ​​for all points in the predefined time interval lies within a specific, predefined tolerance range, the determined pressure profile is considered plausible. The monitoring criterion is met. However, if the deviation of a determined pressure value lies outside the tolerance range, the determined pressure profile is considered implausible, and the respective signal is blocked.In this case, the monitoring criterion is not met. This means that the determined pressure profile or the respective sound signal from the structure-borne sound sensor 10 is considered not sufficiently reliable and is not used for further processing (is blocked).

[0144] In the exemplary embodiment of the Fig. In step S33, the pressure profiles during the period from intake valve closure to ignition point are compared. However, the timing can also be adjusted. For example, based on this pressure, an expected target pressure at the ignition point can be determined using thermodynamics (specifically a polytropic process). In compression-ignition engines, the point in time of the first injection can represent the end of the crankshaft interval under consideration.

[0145] In the exemplary embodiment of the Fig. 4. The pressure profiles are compared by comparing each value of the data-driven calculated pressure profile with the corresponding value of the pressure profile determined by the structure-borne sound sensor 10 within the specified time interval. Alternatively, statistical measures such as the mean or standard deviation of the two pressure profiles can be calculated and compared. If the statistical measures are similar or within a certain tolerance range, the comparison is considered plausible. Otherwise, the comparison is considered implausible. The specific method and tolerance ranges for the comparison can be defined according to the application and specific requirements.

[0146] In the exemplary embodiment of the Fig. 4. Several process parameters are compared over a predefined time interval and monitored using an error threshold to determine the plausibility of the calculated pressure profile. In alternative embodiments, instead of considering a time interval, only the pressure values ​​at a single point in time can be considered. For example, monitoring / comparison only at the time of ignition initiation is one option to define the error threshold for a specific point in time within the considered operating cycle. Tolerance range of the measured pressure curve

[0147] The second aspect of the diagnostic procedure is Fig. 4. If the deviations between the measured pressure values ​​and the corresponding calculated pressure values ​​lie within a specific, predefined tolerance range, the measured pressure profile is considered plausible. The monitoring criterion is then met.

[0148] The diagnostic procedure can include the first aspect described above and / or the second aspect. A combined application of the first and second aspects further improves the traceability of a potential error in a processing chain of the raw signal.

[0149] Fig. 5a and Fig. Figure 5b shows possible embodiments for implementing the tolerance range by means of error thresholds for the second aspect of the diagnostic procedure.

[0150] Fig. 5a and Fig. Figures 5b and 70 each show a visualization of the pressure profile of a cylinder in separate diagrams. The pressure profile represents a value of the internal pressure of a cylinder (axis A1) versus a crank angle (axis A2). A curve of the determined pressure profile (63, 73) (solid curve) and a curve of the corresponding calculated pressure profile (64, 74) (dotted curve) are superimposed.

[0151] The determined pressure profile 63; 73 serves as the output value in step S31 in Fig. 4, in which it is determined from the sound signal of the structure-borne sound sensor 10. The calculated pressure profile 64; 74 serves as the output value in step S32, in which it is calculated from the data-driven model.

[0152] Fig. Figure 5a shows a first embodiment 60 for defining a tolerance range 68. Fig. Figure 5a shows a single determined pressure curve 63 and a single calculated pressure curve 64. The determined pressure curve 63 and the calculated pressure curve 64 have an essentially overlapping profile. Each value of the crank angle on axis A2 is assigned an upper error threshold 65 and a lower error threshold 66. The calculated pressure curve 64 is encompassed by the upper error threshold 65 and the lower error threshold 66. The distance between the upper error threshold 65 and the lower error threshold 66 defines the tolerance range 68. To apply the error thresholds 65 and 66 to the comparison, the deviation between the two pressure curves 63 and 64 is calculated as described in step 33. This deviation is then compared with the error thresholds 65 and 66.As long as the determined pressure profile 63 remains within this tolerance range 68, the monitoring criterion is met, the determined pressure profile 63 is assessed as plausible by the monitoring system, and the determined pressure profile 63 is used (released) for generating the emission prediction. If the determined pressure profile 63 leaves the tolerance range 68, the monitoring criterion is not met, and the sound signal from the structure-borne sound sensor 10 is not used (blocked) for generating the emission prediction. See section 67 in . Fig. Figure 5a visually shows the case where the sound signal is blocked.

[0153] Fig. Figure 5b shows a second embodiment 70 for defining a tolerance range 78. The second embodiment is an alternative or supplement for defining (defining) the tolerance range 78. In contrast to Fig. 5a are in Fig. 5b Several determined pressure curves 73 and several calculated pressure curves 74 are shown. There are sections in the pressure curve where different cylinder pressures are possible at a specific crank angle. This information is known and stored beforehand. A maximum possible value of the cylinder pressure forms the upper error threshold 75, and a minimum possible value of the cylinder pressure forms the lower error threshold 76. The range between the upper error threshold 75 and the lower error threshold 76 defines the tolerance range 78. The plausibility check of the determined pressure curve 73 is carried out as previously described in Fig. 5a explains. Section 77 in Fig. Figure 5b visually shows the case where the sound signal is blocked.

[0154] The first and second embodiments for defining the tolerance range can be used individually or in combination. Setting the error threshold:

[0155] The error thresholds 65, 66; 75, 76 are known in advance. These thresholds represent limits that specify the permissible deviation between the two pressure curves 63, 64; 73, 74 for the determined pressure curve 63; 73 and the corresponding sound signal to still be considered plausible. The specific values ​​for the upper and lower error thresholds 65, 66; 75, 76 are determined based on the specific requirements of the application. They should be chosen to provide adequate tolerance for normal variations and measurement errors while ensuring reliable and accurate results. Various input variables can be used to determine the error thresholds 65, 66; 75, 76. For example, at least one of the following input variables can be included: a legal emission regulation, the engine's mileage to date, or a maintenance check of the engine.The input variables can be stored in a memory. Applying the error thresholds to the comparison in step 33 makes it possible to assess the plausibility of the two pressure profiles and to ensure that only high-quality and reliable data are used to generate the emission prediction. Data-driven model for determining the expected pressure profile:

[0156] In step 32 of the procedure according to Fig. In step 4, a data-driven model of the pressure profile is created. This data-driven model is based on historical and current operating data of the internal combustion engine, collected during normal operation. However, this data is explicitly not the signals from the structure-borne sound sensor used to measure the actual pressure profile in step S31. Instead, the model uses other relevant operating data to generate a reliable prediction of the pressure profile. Data-driven models for calculating and predicting the pressure profile in internal combustion engines are already familiar to experts and are regularly used in engine development and monitoring. These models use historical and current operating data to identify patterns and relationships that are useful for predicting the pressure profile in the cylinders.Such models can be based on various statistical, machine learning or artificial intelligence techniques, including but not limited to neural networks, regression analysis and other algorithmic approaches.

[0157] The specific data used to calculate the pressure profile using a data-driven model can vary depending on the application and available sensors. Operating data used in the data-driven model to calculate the pressure profile in internal combustion engines can be diverse and depend on the specific requirements and context of the model. This includes, for example, data providing information about the engine's operating state, such as engine speed, injection timing, air-fuel ratio, ignition timing, and other relevant parameters. Operating data from other processes or components can also be used. These other processes might include methods for determining ignition timing, camshaft position, or valve position.This data is used in the data-driven model to learn the relationship between the recorded data and the pressure profile in the cylinder. Monitoring the calculated emission forecast by comparison with emission sensor data - Third aspect

[0158] In the first few seconds of an internal combustion engine's operation, when the highest pollutant emissions occur, a NOx sensor, due to its long warm-up time of, for example, one minute, is not yet able to provide reliable data. However, the emission prediction method can already generate a prediction of emissions based on the structure-borne sound signal during this phase. As soon as the NOx sensor is operational or in a transitional phase to provide reliable data, this data can be compared with the emission prediction and calibrated against a confidence interval. This comparison also allows for monitoring and verification of the functionality of the emission prediction method. The comparison can demonstrate that the emission prediction method is functioning correctly.This comparison and alignment corresponds to a comparison with a monitoring criterion.

[0159] In step S31, the pressure profile is determined. The pressure profile in the internal combustion engine 20 during one crankshaft revolution is calculated according to stage ST3 of the previously described multi-stage procedure. Fig. 1

[0160] Fig. Figure 6 shows an example flowchart for the third aspect of the diagnostic procedure. In step S41 of the procedure, the emission forecast (prediction of the pollutant concentration) is generated. The emission forecast is prepared according to stage ST4 of the previously described multi-stage procedure. Fig. 1. The emission forecast is generated even before the NOx sensor is operational. The emission forecast is generated during the start-up phase and the transition range of the internal combustion engine 20, in which the NOx sensor is not yet operational. The generated emission forecast can include a specific modeled pollutant concentration in the exhaust gas.

[0161] In step S42 of the process, the actual pollutant concentration in the exhaust gas is measured using the NOx sensor. The measurement begins at a point in time when the NOx sensor can provide reliable data. The exact time depends on the specific warm-up time of the NOx sensor, which is usually specified by the manufacturer. In the present embodiment, the measurement of the actual pollutant concentration would begin shortly before the NOx sensor provides reliable data. This allows the transition period until the NOx sensor provides reliable data to be mapped and monitored.

[0162] In the subsequent step S43 of the procedure, the generated (modeled) and measured pollutant concentrations are compared. The generated pollutant concentration is compared with the actually measured pollutant concentration. This takes place in the transition range until the NOx sensor is fully operational and functioning correctly.

[0163] In step S44 of the procedure, the confidence interval is determined: Based on a known measurement uncertainty of the sensor and the model uncertainty of the emission prediction generation, a confidence interval is defined. This determination is carried out by combining the measurement uncertainty of the sensor and the model uncertainty of the emission prediction method. This confidence interval represents the smallest possible error threshold for monitoring the modeled emission values.

[0164] The sensor's measurement uncertainty is provided, for example, by the sensor manufacturer and / or can also be determined through calibration and testing. The model uncertainty of the emission prediction method refers to the uncertainties in the modeling and prediction of emissions. It can be determined by comparing the modeled predictions with actual measurements and can be quantified using statistical methods.

[0165] To determine the confidence interval, measurement uncertainty and model uncertainty are combined. This can be done using various mathematical or statistical methods, depending on the nature of the uncertainties and the specific requirements of the diagnostic procedure. A common method is the sum of squares, where the squared uncertainties are added together and then the square root is taken. This yields a total value for the combined uncertainty, which can be defined as the confidence interval.

[0166] In step S45 of the procedure, the comparison result is checked. If the difference between the modeled (generated) and the measured pollutant concentration exceeds the defined confidence interval, the monitoring criterion is not met and a fault code is triggered, for example, an OBD2 fault code. Alternatively or additionally, this can also lead to the triggering of a legally regulated customer restriction, analogous to the SCR system in modern diesel engines. Based on the comparison result, the accuracy and reliability of the emission prediction method can be validated and, if necessary, corrected.

[0167] The comparison in step S45 is performed by calculating the difference between the two values ​​and comparing it to the defined confidence interval. In one embodiment, the comparison can be performed as follows: First, the difference between the modeled pollutant concentration and the actually measured pollutant concentration is calculated. This difference represents the error or deviation between the prediction and reality. The calculated difference is then compared to the previously determined confidence interval. If the difference lies within the confidence interval, the modeled pollutant concentration is considered plausible. The confidence criterion is met.

[0168] The NOx sensor data used for comparison can be generated from point measurements.

[0169] The sensor's measurement uncertainty can be provided by the sensor manufacturer and is found in the sensor's technical specifications. These specifications contain information about the sensor's accuracy, precision, and other performance characteristics.

[0170] The third aspect of the diagnostic procedure described above can serve as an optional supplement to the first and / or second aspects. This third aspect allows for targeted monitoring of the transition period until the emission sensor is fully operational and the combustion engine's pollutant emissions stabilize after a start-up phase. This diagnostic monitoring during the transition period enables verification of the reliability of the multi-stage emission prediction method based on the sound signal. Diagnostic system

[0171] The exemplary embodiments above have demonstrated aspects of a diagnostic procedure applied to emission prediction from a sound signal, which aims to predict and control the emissions of internal combustion engines in real time. In emission prediction (see...) Fig. 1 and corresponding description) the structure-borne sound sensor of an internal combustion engine is used to detect sound signals generated during the operation of the engine. These signals are then analyzed and processed to enable accurate emission prediction. The reliability of the emission prediction method along the processing chain can be demonstrated using the aspects of the diagnostic procedure outlined above.

[0172] Fig. Figure 7 shows three aspects of the diagnostic procedure that are used in the emission prediction procedure. Fig. 1. can be applied. Fig. Figure 7 shows an embodiment in which all three aspects of the diagnostic procedure described above are applied. Applying several aspects of the diagnostic procedure simultaneously can increase the reliability of the emission predictive safety method.

[0173] Fig. Figure 7 shows a first embodiment in which all three aspects of the diagnostic procedure are applied.

[0174] In step S61, the raw signal from the structure-borne sound sensor is monitored and analyzed (see Fig. 2 and corresponding description). Step S61 essentially corresponds to the first aspect of the diagnostic procedure from Fig. 2. Here, the sound signal acquired by the structure-borne sound sensor is transformed into the frequency domain, and a set of relevant frequencies for generating the pressure profile is defined. All frequencies that are unusual in the current operating state of the internal combustion engine 20 limit the confidence value. Step S61 can be used as a step between the first stage ST1 and the second stage ST2 of the emission prediction procedure. Fig. Step S61 can be applied either at or after the second stage ST2 (shown as a dashed line). The placement of step S61 depends on the specific implementation of the first aspect within the multi-stage process. Fig. 1 from.

[0175] In step S62 (see Fig. 4 and corresponding description) the determined cylinder pressure curve is monitored for plausibility. Step S62 essentially corresponds to the second aspect of the diagnostic procedure from Fig. 4. In this aspect, the pressure profile of the internal combustion engine 20, determined on the basis of the sound signal, is continuously monitored and compared with a predicted, data-driven model. Step S62 can be chosen between the third stage ST3 and the fourth stage ST4 of the emission prediction procedure. Fig. 1. can be applied.

[0176] In step S63 (see Fig. 6 and corresponding description) the generated emission prediction is monitored by comparison with emission sensor data. Step S63 essentially corresponds to the third aspect of the diagnostic procedure from Fig. 6. In the third aspect of the diagnostic procedure, the generated emission prediction is compared with the actual measurement data from an emission sensor array, in particular a NOx sensor. Step S63 can be derived from the fourth stage ST4 of the emission prediction procedure. Fig. Step 1 can be applied. This comparison allows the accuracy of the emission forecast to be validated. Alternatively, step S63 can be applied after the fourth stage, ST4. The order of step S63 depends on the specific implementation of the third aspect in the multi-stage process. Fig. 1 from.

[0177] In a second embodiment, the first aspect of the diagnostic procedure is applied in the emission prediction procedure.

[0178] In a third embodiment, the second aspect of the diagnostic procedure is applied in the emission prediction procedure.

[0179] In a fourth embodiment, the first and second aspects of the diagnostic procedure are applied in the emission prediction procedure.

[0180] In a fifth embodiment, the first and third aspects of the diagnostic procedure are applied in the emission prediction procedure.

[0181] In a sixth embodiment, the second and third aspects of the diagnostic procedure are applied in the emission prediction procedure. implementation

[0182] Fig. Figure 8 shows an exemplary representation of a control unit 1200 for monitoring an emission forecast using a multi-stage emission model based on a sound signal from a structure-borne sound sensor.

[0183] Memory 1202 stores various pieces of information, including the software (programs) executed by CPU 1201, as well as data processed during software execution. In particular, memory 1202 stores the program that implements the functions of the control unit 1200 according to the invention. RAM 1203 serves as working memory for CPU 1201 and stores temporary data required during program execution. The accelerator pedal 1210 is an input device that controls the vehicle's speed and acceleration. The position of the accelerator pedal 1210 is read by CPU 1201 and used as an input value for calculating (generating) the emission prediction. The engine interface 1211 serves as an interface to the internal combustion engine 10 of the motor vehicle 100 and enables CPU 1201 to control and monitor the operation of the internal combustion engine 20.The engine interface 1211 can receive various types of information from the internal combustion engine 20, including speed, temperature, pressure, and other operating data. This information is used by the CPU 1201 to monitor the condition of the internal combustion engine 20 and to calculate emission predictions. The structure-borne sound sensor assembly 1212 (structure-borne sound sensor 10) detects sound signals generated by the internal combustion engine 20 during operation. These sound signals are read by the CPU 1201 and used as input values ​​for calculating emission predictions. The user interface 1213 enables interaction between the control unit 1200 and the driver or other vehicle users. The user interface 1213 can include various types of input and output devices, such as screens, touchscreens, gauges, and others.The CPU 1201 can output information and warnings to the user via the user interface 1213, for example, if the measured emissions exceed a certain threshold. The control unit 1200 is programmed to perform the diagnostic procedure described above. The CPU 1201 executes the program stored in memory 1202 to carry out the various steps of the diagnostic procedure, including monitoring and analyzing the sound signal from the structure-borne sound sensor 20, monitoring the plausibility of the measured pressure curve, and monitoring the calculated emission prediction by comparing it with measured emission sensor data from a corresponding emission sensor arrangement 1214. The control unit 1200 can be installed in the motor vehicle 100 according to the invention and used during normal operation of the motor vehicle 100 to monitor and control emissions in real time.Furthermore, the control unit 1200 can also be used in a workshop or test laboratory to monitor emission prediction. The functions implemented in the control unit 1200 can help improve the performance and efficiency of the internal combustion engine 20, reduce emissions, extend the service life of the internal combustion engine, and save costs. The control unit 1200 can also help ensure compliance with emission standards and minimize the vehicle's environmental impact. The control unit 1200 can also help to sufficiently demonstrate the reliability of the emission prediction of the multi-stage emission model based on the sound signal from the structure-borne sound sensor 10. This enables more precise monitoring and control of emissions directly during vehicle operation without the need for invasive and expensive sensor technology for directly measuring exhaust gas composition.

[0184] In a motor vehicle, the 1200 control unit is typically implemented as part of the engine control unit (also known as the Motor Control Unit, MCU, or Engine Control Unit, ECU). The ECU is a specific type of embedded system designed to control and monitor the various systems and subsystems of a motor vehicle.

[0185] Fig.Figure 9 shows a motor vehicle 100 according to the invention, which is equipped with the control unit 110 according to the invention. The control unit 110 can correspond to the control unit 1200. Such a motor vehicle 100 can, by using the diagnostic method according to the invention, monitor an emission prediction in a cylinder of the internal combustion engine 20 based on a sound signal from a structure-borne sound sensor 10 and prevent the generation of the emission prediction based on the monitored sound signal if a corresponding monitoring criterion is not met. This allows the reliability of the emission prediction to be demonstrated.

[0186] The invention described here can be applied in various types of internal combustion engines, including gasoline engines, diesel engines, hybrid engines and others.

[0187] The number of cylinders can be arbitrary. It can be used in various types of vehicles, including passenger cars, trucks, buses, motorcycles, and others. The invention can also be used in stationary applications, such as generators or industrial plants that use internal combustion engines. Reference symbol list 10 Structure-borne sound sensor 20 Internal combustion engine 50 Diagram - Transformed Frequency Band 100 motor vehicles 60 Pressure profile diagram 63 Determined pressure profile 65 Upper error threshold 66 Lower error threshold 68 Tolerance range 70 Pressure profile diagram 73 Determined pressure profile 75 Upper error threshold 76 Lower error threshold 78 Tolerance range 100 motor vehicles 110 Control unit 1200 control unit 1201 Central Processing Unit 1202 storage 1203 RAM 1210 Accelerator pedal 1211 Motor interface 1212 Structure-borne sound sensor arrangement 1213 User interface 1214 Emission sensor arrangement QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2007 007 641 A1

[0005] US 2017 / 0051699 A1

[0006] US 2017 / 0037798 A1

[0007]

Claims

[1] Control unit for an internal combustion engine (20), configured to monitor an emission prediction in a cylinder of the internal combustion engine (20) in real time, which is generated by means of a multi-stage emission model based on a sound signal from a structure-borne sound sensor (10), comprising: - Monitoring a stage within the multi-stage emission model to determine a pressure profile in the cylinder of the internal combustion engine (20) based on the sound signal and comparison with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal if the corresponding monitoring criterion is not met. [2] Control unit according to claim 1, wherein monitoring a stage comprises monitoring the sound signal obtained from the structure-borne sound sensor (10) and the comparison with the monitoring criterion comprises a comparison between a confidence value of the sound signal and an error threshold. [3] Control unit according to claim 1 or 2, further comprising transforming the acquired sound signal from the time domain into the frequency domain into an unfiltered frequency band. [4] Control unit according to claim 3, further comprising defining relevant frequencies that are relevant for determining the cylinder pressure profile and filtering the unfiltered frequency band with a bandpass filter arrangement based on the relevant frequencies into a filtered frequency band. [5] Control unit according to claim 4, further comprising determining a noise component of the sound signal based on a comparison between the unfiltered frequency band and the filtered frequency band and determining the confidence value of the sound signal based on the noise component. [6] Control unit according to one of the preceding claims, wherein monitoring a stage comprises monitoring the pressure profile determined on the basis of the sound signal for plausibility, and the comparison with the monitoring criterion comprises a comparison between the determined pressure profile and a pressure profile calculated from a data-driven model, and a comparison of the comparison with a tolerance range. [7] Control unit according to claim 6, further comprising defining the tolerance range including an upper error threshold and a lower error threshold, wherein the monitoring for plausibility is carried out by comparing the result with the tolerance range. [8] Control unit according to claim 6 or 7, wherein the comparison of the determined pressure profile based on the sound signal with the calculated pressure profile based on the data-driven model is carried out within a predefined time interval. [9] Control unit according to one of claims 6 to 8, wherein determining the pressure profile based on the sound signal and calculating the cylinder pressure profile based on the data-driven model each involve different data. [10] Control unit according to one of the preceding claims, wherein monitoring a stage comprises monitoring the generated emission prediction, and the comparison with the monitoring criterion comprises a comparison between the generated emission prediction and a measured value from an emission sensor and a comparison of the comparison with a confidence interval. [11] Control unit according to claim 10, wherein the comparison between the generated emission prediction and the measured value of the emission sensor takes place in a transition range to an operational readiness of the emission sensor. [12] Control unit according to claim 10 or 11, further comprising issuing an error message if the monitoring criterion is not met. [13] Diagnostic method for monitoring an emission prediction in a cylinder of an internal combustion engine (20) in real time, which is generated by means of a multi-stage emission model based on a sound signal from a structure-borne sound sensor (10), comprising: - Monitoring a stage for determining a cylinder pressure profile in the cylinder of the internal combustion engine (20) based on the sound signal within the multi-stage emission model and comparing it with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal if the corresponding monitoring criterion is not met. [14] Computer program for monitoring an emission prediction in a cylinder of an internal combustion engine (20) in real time, which, when executed by a computer, causes the computer to perform the following procedure: - Monitoring a stage for determining a cylinder pressure profile in the cylinder of the internal combustion engine (20) based on the sound signal within the multi-stage emission model and comparison with a monitoring criterion; and - Preventing the generation of emission predictions based on the monitored sound signal if the corresponding monitoring criterion is not met. [15] Diagnostic system, for an exhaust gas monitoring system for real-time monitoring of a combustion process, for monitoring the exhaust gas monitoring system, comprising: - Control unit according to one of claims 1 to 12; and - a structure-borne sound sensor (10). [16] Motor vehicle (100) comprising an internal combustion engine (20), an exhaust gas monitoring system and a diagnostic system according to claim 15.

Images