Methods and systems for error detection in a production process

Abstract

The present disclosure relates to a computer-implemented method (100) for defect detection in a production process, comprising: - Providing (110) production-related data of a product; - Generating (120), by at least one algorithm, a fault network that specifies a relationship between faults and fault causes, based on the production-related data; and - Determining (130) possible causes of a failure in the production process and probabilities for the possible causes of the failure based on the failure network.

Description

[0001] The present disclosure relates to a method for error detection in a production process, a storage medium for executing the method, and a system for error detection in a production process. The present disclosure relates in particular to error analysis and troubleshooting in vehicle production, such as in the battery cell production of an electric or hybrid vehicle. State of the art

[0002] Hybrid or electric vehicles are powered by an electric motor, with the necessary electrical energy stored, for example, in a high-voltage battery. The high-voltage battery can be charged at a home charging station or a dedicated charging station and represents an energy storage system that stores electrical energy in the form of high-voltage direct current. This stored energy is used to power the vehicle's electric motor. Typically, these high-voltage batteries are composed of lithium-ion cells or modules configured in a battery pack.

[0003] Electrodes are a key component of such high-voltage storage systems. They play a crucial role in the overall weight, peak power, fast-charging capability, and lifespan of the high-voltage storage system. At the same time, electrode manufacturing is characterized by high production reject rates resulting from complex defect patterns that occur at various stages of the process. To precisely identify the causes of these defects, systems are needed that provide targeted and interactive support to technicians, quality managers, and machine operators in troubleshooting. These systems should not only analyze defect patterns but also offer opportunities for process optimization to minimize future rejects and increase production efficiency. Disclosure of the invention

[0004] The purpose of this disclosure is to provide a method for error detection in a production process, a storage medium for executing the method, and a system for error detection in a production process, all of which enable reliable error detection in a production process. In particular, the purpose of this disclosure is to detect and correct errors in a production process.

[0005] This problem is solved by the subject matter of the independent claims. Advantageous embodiments are specified in the dependent claims.

[0006] According to an independent aspect of the present disclosure, a computer-implemented method for defect detection in a production process is specified. The method comprises providing production-related data of a product; generating, by at least one algorithm, a defect network that specifies a relationship between defects and defect causes based on the production-related data; and determining possible defect causes for a defect in the production process (e.g., detected by an employee or automatically) and probabilities for the possible defect causes based on the defect network.

[0007] According to the invention, production-related data is used to construct a fault network that indicates the relationship between faults and their causes. This fault network enables the quantitative calculation of probabilities for potential fault causes as well as the dynamic consideration of new fault patterns. In other words, the present method provides a quantitative method with a dynamic character, in contrast to Failure Mode and Effects Analysis (FMEA), which is a qualitative method with a static character. As a result, reliable fault detection and correction are possible in a production process.

[0008] The term "fault network," as used in this disclosure, refers to a structured model used to systematically represent the relationships between different faults (symptoms) and their possible root causes. In such a fault network, the nodes are represented by faults (symptoms) and possible root causes. The connections between the nodes represent the causal relationships whereby a particular fault is caused by a specific root cause.

[0009] Errors in battery cell production can include, but are not limited to, defects in electrodes, uneven coatings, contamination, or mechanical damage.

[0010] Causes of errors in battery cell production can include, but are not limited to, insufficient material quality, deviations in production parameters, faulty machine calibration or environmental conditions.

[0011] Preferably, the production-related data includes start-of-line (BoL) data. BoL data refers to the information, measurements, and / or quality data that are captured at the beginning of the production process. This data typically originates from the initial stages of production and is crucial for monitoring the condition of raw materials, machine calibration, production equipment parameters, and other relevant factors before the main part of production begins.

[0012] Examples of BoL data for battery cell production include, but are not limited to, material quality (e.g., data on the purity and composition of the raw materials used in electrode manufacturing), machine settings (e.g., parameters such as temperature, pressure, and speed of the machines at the start of production), and environmental conditions (e.g., measurements of temperature, humidity, and other environmental factors that may affect the production process).

[0013] Additionally or alternatively, production-related data includes end-of-line (EoL) data. EoL data refers to the information, measurements, and / or quality data collected at the end of the production process. This data provides insights into the condition and quality of the finished product after all production steps have been completed. EoL data is used to ensure that the final product meets the specified quality requirements and to ultimately verify the success of the production process.

[0014] Exemplary EoL data for battery cell production include, but are not limited to, electrical performance (e.g., measurements of capacity, voltage, and internal resistance of the finished battery cells) and tightness (e.g., results of tests to ensure that the battery cells do not have any leaks).

[0015] Additionally or alternatively, production-related data includes production process data. This data comprises information, measurements, and / or quality data collected during the ongoing production process. Specifically, it relates to the time and / or processes between the BoL (break-by-leave) data and the EoL (end-of-leave) data.

[0016] Exemplary production process data for battery cell production include, but are not limited to, coating parameters (e.g., data on the thickness, uniformity, and composition of the electrode coating during manufacturing) and electrolyte filling (e.g., data on the quantity and distribution of the electrolyte that is filled into the cells).

[0017] Preferably, the at least one algorithm comprises an expectation-maximization algorithm. The expectation-maximization algorithm can be configured for an initial setup of the (initial) error network and optionally for imputing or estimating missing values, e.g., in the production-related data.

[0018] The expectation-maximization algorithm (EM algorithm) is an iterative optimization algorithm used to estimate parameters in models where some data is missing or latent (hidden). According to embodiments of the present disclosure, the strengths of the expectation-maximization algorithm in modeling uncertainties and incomplete data are combined with the ability to structure complex relationships, such as those found in a fault network.

[0019] Preferably, generating the fault network further includes optimizing the (initial) fault network generated by the expectation-maximization algorithm.

[0020] Preferably, generating the error network further includes optimizing the (initial) error network generated by the expectation-maximization algorithm using a genetic algorithm. Genetic algorithms simulate the process of natural selection to improve a population of solutions (often referred to as "individuals") for a given problem over several generations.

[0021] Preferably, generating the fault network further includes parameterizing the (initial) fault network generated by the expectation-maximization algorithm or the fault network optimized by the genetic algorithm. Parameterizing the fault network means providing the connections between the nodes (faults and their potential causes) with specific or local models or parameters that quantify the strength and nature of the relationships.

[0022] Preferably, generating the fault network further comprises performing a parameterization of the (initial) fault network generated by the expectation-maximization algorithm or the fault network optimized by the genetic algorithm using linear regression. In such a parameterization, each relationship between a fault and a cause is described by a linear equation that parameterizes the strength of this relationship.

[0023] Preferably, generating the error network further includes performing a consistency check of the error network. For example, the sequence of process steps and logical connections between relationships can be cross-checked.

[0024] Preferably, the consistency check includes, or is, a Failure Mode and Effects Analysis (FEMA). FEMA is an analytical method in reliability engineering that can provide qualitative statements.

[0025] In this process, potential errors are evaluated, in particular according to their probability of occurrence and their probability of detection, each using a key figure.

[0026] Preferably, the method further comprises controlling or adjusting the production process based on the identified potential causes of errors and their probabilities. For example, at least one production parameter of the production process can be adjusted, set, or changed to prevent the identified error in the future. Examples of production parameters in battery cell production include, but are not limited to, coating speed, coating temperature, layer thickness, calendar printing, etc.

[0027] In one example, the control or adjustment of the production process based on the specific possible causes of errors and the probabilities for those causes can be automated, e.g., by a suitable control system.

[0028] In another example, information on possible causes of errors and / or the probabilities of these causes (e.g., at least part of the error network and the associated probabilities) can be output via a user interface module. Based on this output, a user can control or adjust the production process, or initiate automated control or adjustment of the production process, for example, by means of a corresponding user input at the user interface module.

[0029] The user interface module can include at least one output device and at least one input device.

[0030] The at least one output device can comprise at least one display device and / or at least one loudspeaker. The at least one display device can comprise a display, in particular an LCD display, a plasma display, or an OLED display.

[0031] The at least one input device may include a speech input device and / or a touch-sensitive input device, such as a touch panel or touch pad, and / or a tactile input device, such as a switch (e.g. push button and / or rotary switch) or other mechanically actuated key elements.

[0032] Preferably, the user interface module comprises, or is, a touchscreen that provides at least one output device and at least one input device.

[0033] Preferably, the production process relates to vehicle manufacturing. The term "vehicle manufacturing," as used in this disclosure, refers to the actual manufacturing or production process of vehicles using a manufacturing plant. A vehicle manufacturing plant is a specialized production facility where vehicles and / or their components, such as energy storage devices, are manufactured. These plants may specialize in the production of a particular type of vehicle, such as cars, trucks, or motorcycles, or in specific parts and systems, such as high-voltage storage devices, motors, body parts, or electronic systems. Typical manufacturing plants may include, but are not limited to, energy storage production, assembly lines, press shops, body shops, paint shops, final assembly, quality assurance, and / or combinations thereof.

[0034] The term "vehicle" includes cars, trucks, vans, buses, motorhomes, motorcycles, etc., used for the transport of people, goods, etc. In particular, the term includes motor vehicles for passenger transport.

[0035] Preferably, the product is an energy storage device, such as an energy storage device for an electric or hybrid vehicle. Depending on the embodiment, the hybrid or electric vehicle can be a battery electric vehicle (BEV) or a plug-in hybrid electric vehicle (PHEV).

[0036] The energy storage system in a hybrid or electric vehicle stores electrical energy in the form of high-voltage direct current (DC). This stored energy is used to power at least one of the vehicle's electric motors. Typically, these energy storage systems consist of lithium-ion cells or modules configured in a battery pack. The energy storage system can also be referred to as a high-voltage storage system or battery.

[0037] According to another independent aspect of the present disclosure, a software (SW) program is specified. The SW program can be configured to run on one or more processors and thereby execute the method for error detection in a production process described in this document.

[0038] According to another independent aspect of the present disclosure, a storage medium is specified. The storage medium may include a software program configured to run on one or more processors and thereby execute the method for error detection in a production process described in this document.

[0039] According to another independent aspect of the present disclosure, software with program code is specified. The software is designed to carry out the method for error detection in a production process when the software runs on one or more software-controlled devices.

[0040] According to another independent aspect of the present disclosure, a system for error detection in a production process is specified. The system comprises one or more processors; and at least one memory connected to the one or more processors and containing instructions that can be executed by the one or more processors to perform the error detection method in a production process described in this document.

[0041] A processor or processor module is a programmable computing unit, i.e., a machine or an electronic circuit that controls other elements according to given instructions and thereby advances an algorithm (process).

[0042] Preferably, the system is implemented in a central unit. The term "central unit" refers to a component within a network or system that serves as a control and / or management unit. This central unit can be, for example, a server, a backend, or another comparable device. One of its tasks is to process, store, and / or manage data. Brief description of the drawings

[0043] Examples of the manifestation of the revelation are shown in the figures and are described in more detail below. They show: Fig. 1 a flowchart of a method for defect detection in a production process according to embodiments of the present disclosure, and Fig. 2 a fault detection in battery cell production by means of a fault network according to embodiments of the present disclosure. Implementations of the revelation

[0044] Unless otherwise noted, the same reference symbols are used for identical and equivalent elements in the following.

[0045] Fig. Figure 1 schematically shows a flowchart of a method 100 for defect detection in a production process according to embodiments of the present disclosure. The method 100 can be implemented by suitable software that can be executed by one or more processors (e.g., a CPU).

[0046] The procedure 100 comprises, in block 110, the provision of production-related data of a product; in block 120, the generation, by at least one algorithm, of a fault network that specifies a relationship between faults and fault causes, based on the production-related data; and in block 130, the determination of possible fault causes for a fault in the production process (e.g., detected by an employee or automatically) and of probabilities for the possible fault causes based on the fault network.

[0047] Production-related data is quality-relevant and forms the basis of the defect network. Start-of-line data, production process data, and end-of-line data can be used, but the present disclosure is not limited to these.

[0048] An expectation-maximization algorithm can be used to construct the initial error network and impute missing values. Subsequently, the error network can be optimized using a genetic algorithm, and the relationships within the error network can be parameterized using linear regression.

[0049] The initial fault network generated by the expectation maximization algorithm, the genetic algorithm, and linear regression can then be checked for consistency and meaningfulness. For example, the sequence of process steps and the logical connections in the network can be validated using a Failure Mode and Effects Analysis (FMEA).

[0050] The model can then be applied in real-world operation, meaning that probabilities for potential error causes can be quantitatively calculated. These error probabilities serve as a direct basis for decision-making by temporary workers and / or automated systems. For example, the error causes can be addressed sequentially, starting with the most probable cause, until the error is resolved.

[0051] Fig. Figure 2 shows a fault detection in battery cell production using a fault network according to embodiments of the present disclosure.

[0052] In the upper section of the Fig. Figure 2 shows a fault network for electrode manufacturing in battery cell production. The fault in this case is "electrode surface weight outside tolerance" with the possible causes "collector film surface weight", "slot nozzle pressure outside tolerance", "viscosity outside tolerance" and "mixing ratios outside tolerance".

[0053] In some embodiments, causal chains are possible, meaning that one cause of a fault can lead to another cause of a fault, which in turn causes the fault. In the example of the Fig. 2. “Viscosity out of tolerance” and “Mixing ratios out of tolerance” can each lead to “Slot nozzle pressure out of tolerance”, which in turn leads to the error “Electrode surface weight out of tolerance”.

[0054] In the lower section of the Fig.Figure 2 shows the probabilities associated with the causes of the errors (e.g., "Viscosity is 75% responsible for an excessively high basis weight in the electrode"). These probabilities serve as a direct basis for decision-making by temporary workers and / or automated systems for troubleshooting. For example, the viscosity can be adjusted to bring the slot nozzle pressure within the tolerance range, which in turn brings the basis weight within the tolerance range.

[0055] According to the invention, production-related data is used to construct a fault network that indicates the relationship between faults and their causes. This fault network enables the quantitative calculation of probabilities for potential fault causes as well as the dynamic consideration of new fault patterns. In other words, the present method provides a quantitative method with a dynamic character, in contrast to Failure Mode and Effects Analysis (FMEA), which is a qualitative method with a static character. As a result, reliable fault detection and correction are possible in a production process.

[0056] Although the invention has been further illustrated and explained in detail by means of preferred embodiments, the invention is not limited by the disclosed examples, and other variations can be derived from them by a person skilled in the art without departing from the scope of protection of the invention. It is therefore clear that a multitude of possible variations exist. It is also clear that the embodiments mentioned as examples are truly only examples and are not to be understood in any way as limiting, for example, the scope of protection, the possible applications, or the configuration of the invention.Rather, the preceding description and the description of the figures enable the person skilled in the art to implement the exemplary embodiments in concrete terms, whereby the person skilled in the art, with knowledge of the disclosed inventive concept, can make various changes, for example with regard to the function or the arrangement of individual elements mentioned in an exemplary embodiment, without leaving the scope of protection defined by the claims and their legal equivalents, such as further explanations in the description.

Claims

[1] Computer-implemented method (100) for defect detection in a production process, comprising: - Providing (110) production-related data of a product; - Generating (120), by at least one algorithm, a fault network that specifies a relationship between faults and fault causes, based on the production-related data; and - Determining (130) possible causes of a failure in the production process and probabilities for the possible causes of the failure based on the failure network. [2] Method (100) according to claim 1, wherein the production-related data comprise: - Start-of-line data; and / or - Production process data; and / or - End-of-line data. [3] Method (100) according to claim 1 or 2, wherein the at least one algorithm comprises an expectation maximization algorithm, in particular wherein the expectation maximization algorithm is set up for an initial setup of the error network and an imputation of missing values. [4] Method (100) according to claim 3, wherein generating the fault network further comprises: - Optimizing the error network generated by the expectation maximization algorithm using a genetic algorithm. [5] Method (100) according to claim 3 or 4, wherein generating the fault network further comprises: - Performing a parameterization of the error network generated by the expectation maximization algorithm or the error network optimized by the genetic algorithm using linear regression. [6] Method (100) according to any one of claims 1 to 5, wherein generating the fault network further comprises: - Performing a consistency check of the fault network, in particular by means of a Failure Mode and Effects Analysis. [7] Method (100) according to any one of claims 1 to 6, further comprising: - Controlling or adjusting the production process based on the identified possible causes of failure and the probabilities of those causes. [8] Method (100) according to any one of claims 1 to 7, wherein: - the production process involves vehicle manufacturing; and / or - the product is an energy storage device, in particular an energy storage device for an electric or hybrid vehicle. [9] Storage medium comprising a software program configured to run on one or more processors and thereby to execute the method (100) according to any one of claims 1 to 8. [10] System for error detection in a production process, comprising: one or more processors; and at least one memory connected to the one or more processors and containing instructions that can be executed by the one or more processors to perform the method (100) according to any one of claims 1 to 8.

Images