Quality analysis procedures

Abstract

The present invention provides a computer-implemented method for quality analysis with a quality value, comprising the steps of: recording or acquiring vehicle data that is indicative of a condition of a vehicle or at least one component of the vehicle, recording or acquiring production data of the vehicle and / or of at least one component of the vehicle, wherein the production data is indicative of at least one production parameter, and determining the quality value for quality analysis based on the production data and the vehicle data.

Description

[0001] The present invention relates to a computer-implemented method for quality analysis using a quality value and an analysis system for quality analysis using a quality value.

[0002] During the manufacturing process of vehicle components such as gears, clutches, or shafts, incorrect machine settings and measurement errors can occur. For example, these can include systematic errors that are reproducible under unchanged boundary conditions, such as geometric errors in machine guidance. In addition, random errors can also occur, for example, caused by tool wear or external influences. Errors can lead to a vehicle or component not fulfilling its intended properties, which can lead to complaints or recalls. Errors can also only become apparent and detected during use as abnormal sensor values.

[0003] Establishing a connection between occurring errors and their cause in the manufacture of the vehicle is very complex and often only possible by chance.

[0004] Therefore, it is an object of the present invention to provide a way to analyze errors occurring during use of a vehicle or component.

[0005] The present invention solves the above problem with a method having the features of claim 1, a computer program having the features of claim 10 and an analysis system having the features of claim 11.

[0006] According to one aspect of the present invention, a computer-implemented method for quality analysis with a quality value is provided, comprising the steps of: recording or acquiring vehicle data that is indicative of a condition of a vehicle or at least one component of the vehicle, recording or acquiring production data of the vehicle and / or of the at least one component of the vehicle, wherein the production data is indicative of at least one production parameter, and determining the quality value for quality analysis based on the production data and the vehicle data.

[0007] Compared to the known prior art, the invention is distinguished by the fact that it provides a link between the condition of a vehicle through data acquisition and the production parameters occurring. This provides an indicator for possible anomalies or causes in production. Conclusions can be drawn about the origin of errors. Systematically occurring errors, in particular, can be clarified particularly effectively using the present invention. This allows customer satisfaction, safety, and also production costs to be optimized.

[0008] The computer-implemented method can be a method that can be executed on a computing unit, such as a computer-like device, in particular a computer or a CPU (Central Processing Unit). The computing unit can be designed such that it can perform calculations using a neural network (e.g., deep neural networks, DNN), a convolutional neural network (CNN), and / or a lambda function, etc. The computing unit can preferably be arranged at a central communication point for data. In particular, the computing unit can have access to a comprehensive data memory. Furthermore, a vehicle computing unit can be present in the vehicle, such as power electronics, an on-board computer, or another computing unit intended for automated driving. Furthermore, the vehicle computing unit can also be an additional computing unit.The vehicle data can be recorded and / or acquired by the vehicle computing unit. The vehicle computing unit can be connected to vehicle systems via a CAN bus. The vehicle computing unit can be configured to communicate with the computing unit. Furthermore, a component of the vehicle can also comprise its own vehicle computing unit. This can create independent vehicle modules, such as an axle module, which can communicate independently of the vehicle. The vehicle data can be determinable values ​​and / or variables. The determinable values ​​and / or variables can be determined by the vehicle. In other words, vehicle data can be acquired internally within the vehicle. Furthermore, it is also conceivable for the determinable values ​​and / or variables to be determined using at least one sensor. This can allow a state to be assigned to the vehicle and / or components thereof.The vehicle data can be at least one return value from the at least one sensor. For example, the value can originate from at least one acceleration sensor, a temperature sensor, a rotary encoder, and / or a Hall sensor. The at least one value can be regarded as the vehicle data. The vehicle data can also include at least one value from a slave pointer or a diagnostic counter. This allows vehicle data to be particularly diversified. It is also conceivable for the vehicle data to be determined by the vehicle. Furthermore, vehicle data can also be determined from other vehicle data. In other words, multiple acquired or recorded vehicle data can lead to fewer new vehicle data. This can reduce the computing effort of the computing unit. It is also conceivable for the vehicle data to include information that was not recorded or recorded by the vehicle.In other words, vehicle data can be recorded or captured by at least one external device. Recording of vehicle data by the device can occur, for example, when the vehicle interacts with the device. For example, during a charging process, a charging station can record vehicle data and transmit information to the vehicle. Vehicle data can therefore also include, for example, weather information, timestamps and / or traffic information. This makes it possible to determine a particularly comprehensive state and / or boundary conditions of a vehicle's environment. The vehicle data is, in particular, vehicle field data. In other words, the vehicle data can originate from a large number of vehicles. Furthermore, the vehicle data can originate cumulatively from a large number of vehicles. Vehicle data can be average values ​​across all vehicles in the field.This can enable a particularly balanced determination of an average vehicle condition. Once the vehicle data has been recorded or acquired, the condition can be recorded or determined based on the vehicle data. The condition of a vehicle can comprise the overall condition of all, particularly the essential, components of a vehicle. Essential components can be, for example, the chassis, the drive, the energy storage system, the braking system and / or the infotainment system. The condition can be divided into OK and NIO. The condition can be indicated using consecutive numbers. Furthermore, percentages can also represent the condition. The condition can be indicated using a threshold value. In other words, the condition can be determinable based on the threshold value. Recording or capturing vehicle data can further include consideration of the threshold value.The classification of the status into OK and NIO can depend on the respective threshold value. The vehicle can be an electric vehicle, a hybrid vehicle, a hydrogen-powered vehicle, or the like. In particular, the vehicle can be a passenger car. Furthermore, the vehicle can be designed to drive at least partially automated. In particular, it can be an autonomously operated vehicle. The vehicle can be assigned to one of the autonomy levels 1 to 5. The recording or acquisition of production data can take place in parallel with the recording and acquisition of vehicle data. The recording or acquisition of production data can partially take place before the recording and acquisition of vehicle data. This means that production data can already be analyzed while vehicle data has not yet been recorded or acquired, or has only been recorded in small quantities. There may be more vehicle data than production data.In other words, a number of vehicle data can exceed a number of production data. As a result, more vehicle data than production data can be known. An individual quality value can be determined for each individual vehicle. The production data can include information about the materials used, machines and / or semi-finished products. Furthermore, the production data can include design data such as where production took place. The production data can be recorded or recorded by a production processing unit. A production line or the entire production process can comprise multiple production devices. The production processing unit can in particular be a programmable logic controller (PLC) or a computer-like device, in particular a computer. A production device can have at least one production processing unit.Production data can be transmitted in particular via the production device or via at least one production computing unit. This can include different production units in a manufacturing facility. Production data can include information about production resources. Production data can include information from at least one of the production resources that manufactures or has manufactured the vehicle or components of the vehicle. Production data can include a voltage value, a voltage curve, or a voltage peak. Furthermore, production data can include setting values. This can provide conclusions about possible adjustment work. Production data can be recorded by the production resource and stored in an associated memory. It is also conceivable for the production data to be stored in a central memory, in particular a cloud.This makes it possible to provide cross-system access to the production data. The production data can represent different production parameters. Furthermore, the production data can also specify boundary conditions that prevailed during the production of the vehicle and / or component. For example, the production data can be indicative of climatic boundary conditions (e.g. temperature, humidity, etc.). This makes it possible to use the production data to determine the parameters under which the vehicle or one of its components was processed, manufactured, or manipulated. For example, a result of a tolerance measurement between a gear pair or a shaft-hub connection can be provided as a production parameter through production data. It is also conceivable that a dimension of one of the vehicle components is recorded, in particular manually, after production and entered as a production parameter.The measured dimension of the component can, for example, be stored as associated production data. The quality value can be a measure indicating that a defect, particularly a systematic defect, has occurred during production. In other words, the quality value can be an indicator for quality analysis. This can define a decision variable to initiate an investigation or change to a manufacturing process, or to continue production unchanged. Production data and vehicle data are used to determine the quality value. In other words, the quality value can be dependent on the production data and the vehicle data.

[0009] Furthermore, the quality value can provide a relationship between production data and vehicle data. Specifically, conspicuous, particularly faulty or abnormal vehicle data can be assigned to production data. This makes it possible to determine under which boundary conditions and / or with which settings the specific component or vehicle was manufactured. Such a connection between the vehicle data and the production data can be characterized by the quality value. This makes it possible to identify a production step in which, for example, the most errors occur during subsequent operation of the component or vehicle. This production step can correspond to the vehicle or component from which the respective, particularly faulty or abnormal, vehicle data originate. The quality value can therefore be used to provide conclusions about how vehicle data and production data correlate with each other.It is also conceivable that the quality value can be stored in the central storage, particularly the cloud. This can provide cross-system access to the production data. It is also conceivable that the quality value can provide information for product development and product enhancement. The quality value can, for example, indicate which vehicle or component properties or which materials could be changed to improve the quality of the components. The vehicle data can be recorded directly during operation of the vehicle or the vehicle component. The vehicle data can be measured information. For example, the vehicle data can be indicative of a condition of the vehicle or at least of a vehicle component by being evaluated. Such an evaluation can be carried out by comparison with a limit value.For example, the vehicle data of a transmission (as an example of a component of a vehicle) can include the temperature of the transmission and / or a lubricant in the transmission. As long as the temperature remains below a threshold, the condition of the transmission can be described as OK. As soon as the threshold is exceeded, the condition of the transmission can be changed to NIO. Furthermore, the vehicle data can also be retrieved from a database and thus recorded. This enables subsequent examination of vehicles or vehicle components. Analogously, the production data can be recorded directly during production of a vehicle or a component of a vehicle. In other words, the production data can be recorded in parallel with the manufacture of the vehicle or the component of the vehicle. Alternatively or additionally, the production data can also be recorded by a central storage unit.A production parameter can be indicative of a boundary condition under which the vehicle or vehicle component was manufactured. Furthermore, the production parameter can include the settings of a production facility. If the production data and the vehicle data are then combined, a quality value can be determined. A variety of vehicle data and / or a variety of production data can be used as a basis. The quality value can therefore indicate the circumstances of production as well as the occurrences during operation of the vehicle or vehicle component. The vehicle data and the production data can interact in some way to determine the quality value.This can mean that if the production data indicates a particularly poor value (for example with regard to life expectancy), this can be compensated for by a particularly good value in the vehicle data, so that a neutral quality value can be determined. In one example, this could look like this: the production data indicate that a production machine was operating at the edge or outside of a tolerance range. However, the vehicle data may indicate flawless operation of the vehicle or vehicle component, so that the quality value in such a case can be determined as good or satisfactory. The quality value can be compared with a limit value during a quality analysis. For example, the quality value can be a rating ranging from very good to unsatisfactory, as in the example above. Furthermore, the quality value can specify a numerical value.In any case, the quality value can be compared with a threshold value during a quality analysis to determine whether the desired quality has been achieved. For example, a premium vehicle may be designed to achieve only the highest quality. If the desired goal is not met, the quality analysis can be used to trace the cause of the inadequate quality. This provides a method that can holistically examine and analyze the quality of a vehicle or a component within a vehicle.

[0010] Preferably, the method may further comprise: assigning the vehicle data to different classes, each class forming a group of conditions, the quality value being determined based on the classes and the production data. In other words, this method step may comprise classifying vehicle data into different classes. The classification may be performed by the computing unit. A class may, for example, comprise similar vehicle data. This may reduce the number of different vehicle data. For example, a class may be indicative of a specific measured value range. Thus, all vehicle data of a vehicle or a component of a vehicle that indicate a temperature and have a similar value may be assigned to the same class. Vehicle data may also be assigned to more than one class.The quality value can be determined depending on the classes and the production data. This allows the quality value to correspond to a relationship between classes and production data. In other words, the quality value can be determined based on classes of the vehicle data and production data. This allows the input data (namely the vehicle data) to be reduced when determining the quality value. Consequently, the speed of determining the quality value can be increased. Preferably, different vehicle data can also be assigned to a class. In other words, vehicle data that have a similar or the same effect or indication can be assigned to the same class. This allows vehicle data that have a similar or the same influence on the quality value to be summarized. This can further increase the efficiency of the process.

[0011] Preferably, the assignment of vehicle data to different classes can be carried out by a learning algorithm. In other words, the classification can be carried out using Class-based methods. It is conceivable that the Class-based classification comprises regression, in particular linear regression. Other self-learning algorithms and / or a combination of several algorithms are also conceivable. This makes it possible to provide the most adaptive assignment of vehicle data to different classes possible. In particular, in the case where several vehicle data items are to be summarized in one class (i.e., assigned to this class), a learning algorithm delivers good results. For example, such an algorithm can learn using data from the past which vehicle information can be summarized in a class. For example, based on the influence this vehicle data has on the quality value.This can further simplify the procedure and at the same time increase its informative value.

[0012] The method preferably comprises creating new classes based on the vehicle data. In other words, a new class can be created by the learning algorithm if, for example, vehicle data or a combination of vehicle data occurs that cannot be easily assigned to an existing class. Furthermore, new classes can also be created if the resolution of the results is to be increased. New classes can, for example, be created manually by a user or automatically. In particular, new classes can be created by the learning algorithm. This allows the classes to vary depending on the environmental conditions in which the vehicle or vehicle component is used. This ensures that the method can adapt to the respective application situation.

[0013] Preferably, the vehicle data can be classified automatically. In other words, the vehicle data can be processed through a self-executable determination on the computing unit. This can reduce manual effort and manual errors.

[0014] Preferably, the vehicle data can be recorded and / or transmitted while a vehicle is traveling. In other words, the vehicle data can be recorded and / or transmitted while the vehicle or a component of the vehicle is in use. This includes, for example, parking, charging, or servicing the vehicle. This allows vehicle data to be recorded and / or determined in various practically possible operating states. Furthermore, the vehicle can have a data interface. The vehicle can communicate the vehicle data via the data interface. Furthermore, it is conceivable that a vehicle journey also includes, for example, testing the vehicle or a component of the vehicle on a test bench.

[0015] Preferably, the vehicle data can be stored, at least temporarily, on at least one decentralized memory of the vehicle. At least temporarily, a period of time can span from the recording or acquisition of the vehicle data to the communication (i.e., output) of the vehicle data. Communication can be wireless. It is also conceivable that communication could be performed during service and / or repair work.

[0016] Preferably, the vehicle data can be recorded or acquired from multiple vehicles. Vehicle data can be specific to the respective vehicle in which this vehicle data is recorded. In other words, vehicle data for the same components or vehicles can originate from different vehicles. This allows the vehicle data to be diversified particularly advantageously, especially within a class.

[0017] Preferably, the vehicle data can be recorded or acquired, in particular exclusively, when the vehicle or component is in a faulty state. In other words, the vehicle data can only be recorded or acquired once a fault has been detected. A fault can exist, for example, when measured values ​​exceed a set threshold. This can have the effect that the faulty state can be assigned to a possible cause during production of the vehicle or component using quality analysis. The faulty state can be defined as a defect or limited, in particular disrupted, functionality of the vehicle. For example, the faulty state can manifest itself in the recording or acquisition of increased temperatures or increased noise development from a gear pair.Vehicle data in a faulty state may differ from the distribution of the corresponding vehicle data in a normal state, particularly as outliers. It is also conceivable that vehicle data in a faulty state may be used preferentially to determine the quality value. This allows conclusions to be drawn about previous production defects.

[0018] Preferably, the fault condition can occur when an abnormality in vehicle data is detected by a user or automatically. This can occur, for example, when a user indicates via an interface that the vehicle or a component is abnormal, according to the user's perception. Abnormal can mean contrary to expectations or known behavior. In other words, the vehicle or components of the vehicle can be put into a fault condition by driving or usage behavior that is perceptible to the user.

[0019] The classes can preferably include speed intervals, torque intervals, mileage, diagnostic information from display means, and / or vibrations. In other words, at least one class can be defined that, for example, covers a speed range from 0 to 1500 engine revolutions per minute, and another class that covers 1500 to 2500 engine revolutions. As already described above, several different vehicle data can also be assigned to a class. Furthermore, the classes can also be divided according to measurable effects. For example, vehicle data that causes a specific vibration in the vehicle or component of the vehicle can be assigned to a class. For example, it can happen that a vibration in the range above 80 Hz is detected. In this case, vehicle data, in particular that which is the cause of the vibration, can be assigned to a class.In this sense, a variety of classifications are conceivable. This allows for the inclusion of relevant areas in which the vehicle may be located during normal use.

[0020] Preferably, the method can further comprise: determining customer behavior based on the vehicle data. In other words, the vehicle data can be indicative of customer behavior or can be interpreted as such. Customer behavior can be specific to singular or multiple vehicle data. In other words, customer behavior can be defined by a set of vehicle data. Customer behavior can be indicative of the fault condition. In other words, customer behavior can determine whether a fault condition exists. It is also conceivable for customer behavior to be included in the determination of the quality value. Thus, possible fault conditions can only be indicative of conspicuous production data under certain customer behaviors. Furthermore, customer behavior can also indicate how the vehicle or component of the vehicle is used.Heavy usage can occur if, for example, a user drives the vehicle very frequently and / or very quickly. Customer behavior can be taken into account when classifying vehicle data into classes. This allows specific user groups to be considered in isolation and, in particular, the quality value to be determined taking their specific needs into account.

[0021] Preferably, the quality value can comprise a predictive value indicative of a future condition of the component or the vehicle. In other words, the quality value can comprise an influencing factor for predictive maintenance. Based on the quality value, for example, a remaining service life of the vehicle or one of its components can be estimated. Accordingly, the method can additionally comprise the step of determining a remaining service life based on the quality value. This allows maintenance of the component of the vehicle or the vehicle to be planned. Furthermore, the user can be informed in a timely manner about a possible failure of the component of the vehicle or the vehicle.

[0022] Preferably, the production data can include manual inputs from measurement methods. The production data can be recorded or acquired during production. The dimensional accuracy of at least one component of the vehicle can be manually checked. A manual measurement result can then be fed into a recording system by manual input. In other words, production data can be both manually recorded and manually acquired. This has the effect of allowing the production data to be particularly diversified. Furthermore, production data from different manufacturing steps of one of the components can be recorded and acquired.

[0023] Preferably, the production data can be recorded during the manufacture of the vehicle and / or a component thereof. The production data can include individual processing and / or manufacturing steps. Furthermore, the production data can include preset operating values ​​and actually performed processing steps of machines and / or production aids. Production aids can be, for example, torque wrenches, whose tightening torque can be, in particular, production data. This allows conclusions to be drawn from the production process regarding the finished end product.

[0024] Preferably, the vehicle and / or the production device can each have a communication interface, wherein the communication interface provides wireless communication. The communication interface can be configured to receive and / or transmit vehicle data or production data. In particular, the method can comprise the step of communicating the production device and / or the vehicle and / or component with a computing unit. This can result in the vehicle or production data being able to be fed to a data storage device via the computing unit. The determination of the quality value can be enabled by the computing unit accessing the data storage device.

[0025] The method can preferably further comprise: outputting the quality value on a display interface, in particular a dashboard. The display interface can be in contact with the computing unit at the central communication point for data. This means that the computing unit can be designed to control the display interface. The display interface can be a digital interface, in particular a program. It is also conceivable for the display interface to display a list of quality values ​​for individual components of the vehicle. Likewise, the display interface can also display average values, in particular multiple quality values. The display interface can further display information that is the cause of a displayed quality value. Communication via the dashboard can preferably be bidirectional. For example, a user can initiate input data via the dashboard.The user can assign weights to individual vehicle and / or production data and thus examine their impact on the quality value. This can improve the traceability of a specific quality value.

[0026] Preferably, the vehicle and / or the production device can communicate with a central processing unit. The processing unit can correspond to the aforementioned processing unit. The processing unit can be a standalone unit. The processing unit can be configured to record both vehicle data and production data. Furthermore, the processing unit can be configured to determine the condition of a vehicle based on the vehicle data. The processing unit can also be configured to determine the production parameters based on the production data. This allows data to be processed on the processing unit.

[0027] Preferably, the computing unit can access and / or comprise a central storage device, in particular a cloud storage device. The central storage device can be a data storage device. The computing unit can be configured to store the production data and / or vehicle data in the central storage device and load them from it. This allows both vehicle data and production data from multiple interfaces, such as vehicles or production resources, to be made accessible to the computing unit.

[0028] Preferably, the computing unit can be configured to determine the quality value. In other words, the computing unit can be configured to generate the quality value. This allows the quality value to be determined based on centrally collected vehicle data and production data.

[0029] Preferably, the quality value is determined based on the vehicle data, in particular the classes, and the production data using canonical correlation analysis. Canonical correlation analysis is a multivariate statistical method for analyzing the dependence of two random vectors. Furthermore, it allows these relationships to be summarized into a smaller number of statistics for high-dimensional data. In this case, the two vectors can be the vehicle data (or classes) and the production parameters. In other words, canonical correlation analysis can refer to two groups of characteristics that can be formed by the classes and the production parameters. Each of these groups can be a linear combination of weighted variables. Each variable can therefore have its own coefficient or weighting factor. The respective coefficient or weighting factor can be variable.In other words, both a weighted linear combination for the characteristics of the classes and a weighted linear combination for the characteristics of the production parameters can be formed. A relationship between the two linear combinations can be determined by a correlation value. It is conceivable that the quality value corresponds to the correlation value. This advantageously allows a question such as the following to be asked: With what correlation can a specific condition be traced back to a specific production parameter? This advantageously provides an analysis of a relationship between at least one, in particular a defective, vehicle or component condition and at least one production parameter.

[0030] Preferably, the canonical correlation analysis can be performed for a plurality of classes and a plurality of production parameters. In other words, the respective linear combinations can comprise more than just a product of weighting factor and feature. This allows a diverse question to be answered using canonical correlation analysis. In other words, several features of the classes can be correlated with features of the production parameters. In practical and concrete terms, this can mean that a defective chassis (class) can be related to an incorrect tightening torque during vehicle production (production parameter) with a specific correlation.Correlation can also represent the probability of a mutual relationship between two different characteristics, with absolute value correlations from 0 to 0.5 defining a low probability and from 0.5 to 1 defining a higher probability. This can advantageously be used to determine statements from correlation analysis.

[0031] The method preferably further comprises: performing a quality analysis by comparing the quality value with a target variable. The target variable can be a key performance indicator (KPI). The target variable can thus define requirements or goals that are to be achieved, in particular during or through optimization of production. For example, the target variable can be to increase the quality, service life, and / or life expectancy of the vehicle or a component of a vehicle. Accordingly, those production parameters can be determined under respective quality values ​​that can be assigned to those classes associated with defective vehicle values. In other words, the target variable can define a benchmark that is to be achieved. With the present method, the vehicle data and / or production data that are causal for a specific quality value can be determined.Adjustments can then be made, for example, in production. This allows a target value to be achieved in the future. The quality values ​​can then be used to provide quality analysis. Consequently, those production parameters that may be related to quality defects can also be identified.

[0032] Preferably, the target variable can include a remaining service life. In other words, the quality analysis can determine the condition of the vehicle and / or one of its components. It is also conceivable that the quality analysis can provide predictable maintenance of the vehicle or one of its components. This can increase vehicle functionality and driving safety.

[0033] Preferably, the production data can be stored, at least temporarily, on at least one decentralized memory of the production device that receives or records the production data. This can provide the advantage that production data can be communicated multiple times before being deleted. Furthermore, it can ensure that the processing unit receives the production data correctly. The decentralized memory can be a memory assigned to the production device that can communicate with a device processing unit. The memory can be part of the production device. This can provide particularly streamlined communication between the production device and the decentralized memory.

[0034] Preferably, the vehicle can comprise at least one sensor that communicates with a vehicle computing unit. The vehicle computing unit can correspond to the aforementioned vehicle computing unit of the vehicle. The sensor can be configured to detect signals and / or values ​​and send them to the vehicle computing unit. One of the sensors can be configured to measure acceleration, vibration, temperature, voltage, and / or other physical properties. Vehicle data can thus be generated using the at least one sensor.

[0035] Preferably, signals detected by the at least one sensor can be processed by the vehicle computing unit, wherein the signals can be vehicle data. It is also conceivable for the signals and / or values ​​of the at least one sensor to be fed to the computing unit via the vehicle computing unit. The computing unit can also be configured to process the signals in such a way as to obtain vehicle data. This allows for a versatile architecture to be provided.

[0036] According to a further aspect of the invention, a computer program with program code is provided which, when executed on a computer unit, can be designed to carry out the above method. The computer program can be in any code, in particular in a code suitable for analyzing data sets. According to a further aspect, the invention is directed to a computer-readable medium comprising a computer program as defined above. The computer-readable medium can be any digital data storage device, such as a USB stick, a hard disk, a CD-ROM, an SD card, or an SSD card. The computer program does not have to be stored on such a computer-readable medium in order to be made available to a user, but can also be obtained via the Internet.

[0037] According to a further aspect of the invention, a computer unit is provided that can carry out the above method. The computer unit can also be the above-mentioned computing unit. The computer unit can be configured to read and process the program code. Furthermore, the computer unit can be configured to communicate with one or more of the vehicles. The computer unit can also communicate with one or more production computing units of the production devices.

[0038] According to a further aspect of the present invention, an analysis system for quality analysis with a quality value is provided. The analysis system comprises: a vehicle data acquisition device for receiving or acquiring vehicle data that is indicative of a condition of a vehicle or at least one component of the vehicle, a production data acquisition device for receiving or acquiring production data of the vehicle and / or at least one component of the vehicle, wherein the production data is indicative of at least one production parameter, and a computing unit that is designed to determine the quality value for quality analysis based on the production data and the vehicle data. The analysis system can be provided centrally and receive the information, for example, via the Internet. Alternatively or additionally, the analysis system can be located in the vehicle itself or on a component orA central analysis system can be provided as part of the vehicle. In this case, the analysis system can be provided in a decentralized manner. Furthermore, the analysis system can record the vehicle data itself, in particular using its own sensors. It is also conceivable for a central analysis system and a decentralized analysis system to be interconnected. In this case, the decentralized analysis system can perform pre-processing of the vehicle data. For example, the pre-processing can include classifying the vehicle data. This can simplify and accelerate data transmission.

[0039] According to one embodiment of the present invention, product development and product enhancement require information on which properties should be added to the product, which properties need to be changed, which materials or components need to be improved, which materials or components can be replaced with cheaper alternatives, as well as further feedback from customers. Missing information leads to (expensive) incorrect developments and to a lack of (cost) optimization of the materials or components. Through targeted, AI-supported data analysis of vehicle field data (e.g., vehicle data) in which the products are installed, error analyses, clustering of customer groups, and customer behavior analyses can be carried out. The present invention according to one of the above embodiments can provide this and, in addition, offers cost reduction through data-driven product development.

[0040] According to a further embodiment of the present invention, defects can be detected and remedied at an early stage through targeted, AI-supported data analysis of vehicle field data in which the products are installed, in conjunction with quality data from production / manufacturing. Specifically: Field data (i.e., vehicle data), which includes, among other things, trailing indicators and diagnostic counters of error entries, are first divided into classes using AI-supported classification. These can be classes of different intervals of speeds and torques, which in turn reflect driver behavior, as well as classes based on mileage or vehicle characteristics. However, classes can also be formed from the trailing indicators and diagnostic data. For this purpose, a suitable distance measure must be selected to measure similarity.

[0041] With the resulting classes, a canonical correlation analysis can be performed, which determines a redundancy measure that indicates what proportion of the total variance of the canonical variables can be explained by each variable. Canonical correlation analysis refers to two groups of features. This means that instead of one variable Y, p, variables Y1, ..., Yp are linearly combined with coefficients a. k into the analysis, which form a supervariable G or F. For two group variables X, Y this means that for them coefficients a k and b jare to be determined in such a way that the correlation between the supervariables G and F is maximized. The supervariables G and F are not directly observable, but must be determined in the algorithm. This makes it possible to illustrate the intra-domain loading, which defines the relationship between the X variable and the canonical values ​​F or Y and G. And secondly, the inter-domain loadings, which illustrate the relationship between the X variables and the canonical values ​​of G or Y to F. The inter-domain loadings can therefore be used to draw conclusions, e.g. from the diagnostic entries for the respective class variables, and thus to detect and identify malfunctions or component damage at an early stage. This method also supports finding correlations between error patterns in product operation, as well as in production and assembly, and to confirm or reject causal relationships. This can therefore be used toindividual tolerance levels, production and assembly steps are re-evaluated and adjusted.

[0042] According to a further embodiment, an extension of the existing data, such as data from the end-of-line test bench or development data, can improve the possibility of drawing conclusions about the causes.

[0043] Thus, according to one embodiment, targeted, AI-supported data analysis of vehicle field data (vehicle data) in which the products are installed can be used to perform error analyses, clustering of customer groups, and customer behavior analyses. By analyzing the vehicle field data, many questions can be clarified that provide important information for product development, such as how the product should be further developed, or even which new products would be appropriate based on user / customer behavior. Examples of questions that can be clarified using vehicle field data analyses include:: What are the mean values ​​and variances of the forces / stresses on the components and materials in the real world (field data), can the quality tests for future products be adapted to simulate more realistic stress, how high is the frequency and therefore the priority of the various fail-safe states (in the event of component faults and / or which components are stressed significantly below expectations and can therefore be replaced with cheaper components without a reduction in quality. By answering these and many other questions, product development can be improved, for example by choosing better materials, surfaces and components for the questions mentioned. More realistic quality tests can find faults in new products before they even go into series production. By improving prioritization, a lower probability of failure can be achieved.By using lower-cost components, a higher profit margin can be achieved while maintaining the same quality.

[0044] Therefore, the present invention can be used to obtain information required for product development and product enhancement, such as which properties should be added to the product, which properties should be changed, which materials or components need to be improved, which materials or components can be replaced with cheaper alternatives, as well as further customer feedback. A lack of information leads to (expensive) misdevelopments and a lack of (cost) optimization of materials or components. This can be prevented with the present method.

[0045] Individual embodiments and features can be combined with other embodiments and other features to form new embodiments. Embodiments and advantages of the embodiments and features also apply analogously to the new embodiments. Furthermore, embodiments and advantages mentioned in connection with the method also apply analogously to the device, the computer program, and / or the computer unit, and vice versa.

[0046] Embodiments of the present invention will be described in detail below with reference to the accompanying figures. Fig. 1: a schematic representation of devices of the method according to the invention in one embodiment, Fig. 2: a further schematic representation of information flows of the method according to the invention in one embodiment, and Fig. 3: a flowchart of the method according to one aspect of the present invention.

[0047] In the figures, identical features are identified by identical reference numerals.

[0048] Fig. Figure 1 shows a schematic representation of devices of the method according to the invention. The method can also use other devices than those shown in Fig. 1. The computer-implemented method records, on the one hand, data originating from vehicles 10 and from production devices 20. Several vehicles 10 and several production devices 20 can communicate with a central processing unit 1. In order to determine the data, these must first be detected as signals by suitable detection means. Accordingly, a vehicle 10 can comprise at least one sensor (not shown) or other detection device, which in the embodiment shown can communicate with a vehicle processing unit 14. Vehicle data can be formed from the signals of a vehicle 10 by the vehicle processing unit 14. It is also conceivable for the processing unit 1 to receive sensor signals from the vehicle 10 and form vehicle data therefrom. This can reduce the computing effort on the vehicle processing unit. Furthermore, a vehicle 10 comprises a decentralized memory 12.Vehicle data or sensor signals can be stored on the decentralized memory 12. The vehicle 10 can transmit vehicle data or sensor signals to the computing unit 1 via a communication interface of the vehicle 10. Vehicle data can also comprise components of a vehicle 10, in particular dedicated assemblies. These components can be independently designed to communicate vehicle data or sensor signals. The number of vehicles 10 can be one, two, or more. In addition, a production device 20 can also comprise a production computing unit 22. In further embodiments, it is also conceivable for several production devices 20 to comprise a common production computing unit 22. At least one sensor (not shown) or a detection device can likewise be arranged on a production device 20.Production data can be generated by the production processing unit 22 from the signals of a production device 20. It is also conceivable for the processing unit 1 to receive sensor signals from the production device 20 and generate vehicle data from them. This can reduce the computational effort on the production processing unit. The number of production processing units 22 can be one, two, or more.

[0049] Fig. 2 shows a schematic representation of a relevant information flow according to an embodiment of the present invention. The vehicle 10 shown can provide information as vehicle data. The vehicle data can be stored on the decentralized memory 12. From the decentralized memory 12, the vehicle data can be fed to the computing unit 1. The production devices 20 shown can provide information as production data. The production data can be stored via the production computing unit and fed to the computing unit 1. The computing unit 1 can then perform a quality analysis based on both vehicle data and production data. In other words, the quality analysis can receive vehicle data and production data. The quality analysis can include a correlation analysis, wherein a correlation between vehicle data and production data can be determined.Accordingly, relationships between the vehicle data and the production data can be established using a quality value. In other words, the result of the quality analysis can be the quality value. The quality value can be stored, in particular, on the computing unit 1. The quality value can also be fed to a dashboard (not shown) via an interface and output there.

[0050] Fig.3 shows a flowchart of the computer-implemented method. The vehicle data acquisition S1 and the production data acquisition S2 can be carried out simultaneously, in particular in parallel. The vehicle data acquisition S1 can involve a greater scope, in particular more computational effort, than the production data acquisition S2. This can have the effect that more vehicle data is recorded or acquired than production data. The quality analysis S3 can receive vehicle data from the vehicle data acquisition S1 and production data from the production data acquisition S2. The quality analysis S3 can also be carried out multiple times, in particular based on modified and / or further developed vehicle data and / or production data. In other words, the vehicle data acquisition S1 and the production data acquisition S2 can be carried out continuously, while the quality analysis S3 can be carried out at specific times.This allows the quality value to be determined as needed, especially after special developments in the vehicle or production data have occurred. Reference symbol 1 computing unit 2 cloud storage 10 vehicles 12 decentralized storage 14 Vehicle computing unit 20 Production device 22 Production calculation unit S1 vehicle data acquisition S2 Production data acquisition S3 Quality Analysis

Claims

[1] Computer-implemented method for quality analysis with a quality value, comprising the steps: - Recording or capturing vehicle data that is indicative of the condition of a vehicle or at least a component of the vehicle, - Recording or capturing production data of the vehicle and / or at least one component of the vehicle, wherein the production data are indicative of at least one production parameter, and - Determining the quality value for quality analysis based on production data and vehicle data. [2] Method according to claim 1, wherein the method further comprises: assigning the vehicle data into different classes, each class forming a group of states, the quality value being determined based on the classes and the production data. [3] Method according to one of the preceding claims, wherein the assignment of the vehicle data into different classes is carried out by a learning algorithm. [4] Method according to any of the preceding claims, wherein the vehicle data is recorded and / or transmitted during the journey of the vehicle. [5] Method according to any of the preceding claims, wherein the vehicle data are recorded or captured, in particular exclusively, in a fault condition of the vehicle or component. [6] Method according to any of the preceding claims, wherein the method further comprises: determining customer behavior based on the vehicle data. [7] Method according to any of the preceding claims, wherein the quality value includes a predictive value that is indicative of a future state of the component or the vehicle. [8] Method according to one of the preceding claims, wherein the determination of the quality value based on the vehicle data, in particular the classes, and the production data is carried out by a canonical correlation analysis. [9] Procedure according to one of the previous steps, wherein the procedure further comprises: carrying out a quality analysis by comparing the quality value with a target value. [10] Computer program with program code which, when executed on a computer unit, is configured to perform the method according to any of the preceding claims. [11] Analysis system for quality analysis with a quality value, comprising: a vehicle data acquisition device for recording or capturing vehicle data that is indicative of the condition of a vehicle or at least of a component of the vehicle, a production data acquisition device for recording or capturing production data of the vehicle and / or at least one component of the vehicle, wherein the production data are indicative of at least one production parameter, and a computing unit designed to determine the quality value for quality analysis based on production data and vehicle data.

Images