System and method for predicting the expected lifespan of a machine, AI-based prediction module, and method for training an AI-based prediction model.

Abstract

The invention relates to a system (1) for predicting the expected service life of a machine (2), wherein the machine (2) has a sensor module (3) for continuously measuring parameters (4) influencing the service life of the machine (2), wherein the parameters (4) can be transmitted via a transmission module (6) to a central storage module (7) and stored there, wherein the system (1) further comprises an input module (8) for receiving and storing data representing a mechanical model (9) of the machine (2), and a first prediction module (10) which is connected to the central storage module (7) and the input module (8) for determining a first service life prediction (11) of the machine (2) based on the determined parameters (4) and the mechanical model (9) of the machine (2), wherein the system (1) further comprises an output module (12).via which the first lifetime prediction (11) of the first prediction module (10) can be visualized for a user, wherein the central storage module (7) is further connected to an AI-based second prediction module (13) for determining a second lifetime prediction (14) of the machine (2) based on the determined parameters (4), and the second lifetime prediction (14) of the second prediction module (13) can be visualized for a user via the output module (12).

Description

[0001] The present invention relates to a system and a method for predicting an expected service life of a machine, wherein the machine has a sensor module for continuously measuring parameters influencing the service life of the machine, wherein the parameters can be transmitted to a central storage module via a transmission module and stored there, wherein the system further has an input module for receiving and storing data representing a mechanical model of the machine, and a first prediction module which is connected to the central storage module and the input module for determining a first service life prediction of the machine on the basis of the determined parameters and the mechanical model of the machine, wherein the system further comprises an output module via which the first service life prediction of the first prediction module can be visualized for a user.The invention further relates to an AI-based prediction module and a method for training an AI-based prediction model.

[0002] The state of the art in preventive maintenance, and particularly in systems for predicting the service life of machines, has made remarkable progress. Nevertheless, between 15 and 35 percent of machine failures are still due to overload or fatigue. Such failures result in significant downtime costs, especially when critical and expensive machines are involved and spare parts are not readily available. Although methods such as the L10 calculation exist to estimate the expected service life of machine components, these approaches reveal significant accuracy limitations.

[0003] Precisely predicting a machine's service life is a complex challenge, complicated by several factors. On the one hand, existing models are often unable to accurately model real-world stress and the resulting fatigue phenomena under varying operating conditions. On the other hand, many methods do not account for sudden or irregular events that can lead to premature failure. Furthermore, the performance of currently used methods is limited by the fact that they often rely on linear extrapolation of data, leading to inaccurate predictions in dynamic and complex operating environments. In particular, they often lack a holistic integration of real-time operating data and advanced analytical algorithms capable of capturing and modeling the nuanced nature of machine operation.

[0004] One of the critical limitations of existing systems is the inability to develop adaptive models that enable real-time learning from operating data and continuously update the predictive algorithms. Furthermore, many current systems are tailored to specific machine types or operating conditions, offering little flexibility for application to different machines or under changing circumstances. Furthermore, user acceptance is limited by a lack of usability and the challenge of translating complex predictive results into practical maintenance decisions.

[0005] The object of the invention is therefore to avoid or at least reduce the disadvantages known from the prior art and to provide an improved system and an improved method for predicting the expected service life of a machine. Furthermore, the object of the invention is to implement an optimized AI-based prediction module and an optimized method for training an AI-based prediction model.

[0006] This object is achieved by a system for predicting an expected service life of a machine, wherein the machine has a sensor module for continuously measuring parameters influencing the service life of the machine, wherein the parameters can be transmitted via a transmission module to a central storage module and stored there, wherein the system further comprises an input module for receiving and storing data representing a mechanical model of the machine, as well as a first prediction module which is connected to the central storage module and the input module for determining a first service life prediction of the machine on the basis of the determined parameters and the mechanical model of the machine, wherein the system further comprises an output module via which the first service life prediction of the first prediction module can be visualized for a user,wherein the central storage module is further connected to an AI-based second prediction module for determining a second lifetime prediction of the machine based on the determined parameters, and the second lifetime prediction of the second prediction module is visualizable for a user via the output module.

[0007] This provides the advantage of continuously measuring a machine's service-critical parameters in conjunction with a second prediction module based on artificial intelligence (AI). This enables a more precise and dynamic forecast of the expected service life. The implementation of a mechanical model of the machine into the prediction system allows specific characteristics and behaviors of the machine to be incorporated into the forecast, further increasing forecast accuracy. The ability to visualize the prediction results for the user also improves the accessibility and comprehensibility of the forecast information, which can contribute to more efficient planning of maintenance intervals and an extension of the machine's service life.As a result, unplanned downtime can be minimized and operational efficiency increased, bringing both technical and economic benefits.

[0008] For the purposes of this patent application, a system for predicting the expected service life of a machine is a constellation of components and modules that interact to estimate the future service life of a machine based on the analysis and evaluation of parameters that influence the machine's life expectancy. This system forms an integrative solution that enables predictive maintenance decisions, thus optimizing operational efficiency and reducing unplanned downtime.

[0009] The function of this system is to collect, store, process, and analyze data to make predictions about the remaining service life of the machine. The accuracy of these predictions is crucial for optimizing maintenance schedules and minimizing downtime.

[0010] The system structure includes the following components: Sensor module: The sensor module preferably contains sensors for continuously measuring parameters that influence the service life of the machine. These parameters can be mechanical or electrical in nature, such as vibrations, forces, torques, current, and voltage. The sensors are used to collect data in real time and make it available for further processing. Transmission module: The transmission module transmits the collected data to a central storage module. This transmission can be achieved using various communication technologies, with wireless solutions being preferred due to their flexibility and scalability. Central storage module: The central storage module, ideally configured as a cloud storage, serves to centralize the data collected by the sensors. The cloud-based solution offers advantages in terms of accessibility, scalability, and data security. Input module: The input module allows the input and storage of additional data relevant for lifetime prediction. This includes, in particular, a mechanical model of the machine, which represents the physical and structural properties and serves as essential input for the prediction module. Prediction modules: The first prediction module connects the central storage module and the input module to determine an initial service life prediction for the machine. Furthermore, the system is equipped with an AI-based second prediction module that enables a second service life prediction. This AI-based analysis can preferably incorporate both the determined parameters and the mechanical model of the machine. Output module: The output module visualizes the determined lifetime predictions and makes them accessible to the user. The output module can take various forms, such as a PC, a tablet computer, or a smartphone.

[0011] In the context of this application, "continuous measurement" refers to the process of measuring parameters at regular or irregular intervals without significant time gaps, enabling comprehensive and continuous data collection. This includes both continuous real-time monitoring and frequent discrete measurements, e.g., multiple times a day, to ensure a consistent and up-to-date flow of information reflecting the current condition and performance of the machine. This addresses various practical data collection scenarios, whereby continuous data collection can be achieved through high-frequency, periodic measurements that together result in a nearly continuous data set.

[0012] For the purposes of this patent application, a sensor module is a combined unit of one or more sensor types, specifically designed to record relevant machine parameters that can influence its service life and operating condition. This module forms a potential basis for data collection within the overall system for predicting a machine's service life. The function of the sensor module is to continuously record data relevant for assessing the current condition and expected service life of the machine. This includes, for example, measuring physical and electrical parameters such as temperature, pressure, vibration, current, and voltage. These parameters provide information about signs of wear, loads, and potential sources of error in the machine, which are analyzed for preventative maintenance and to avoid unplanned downtime.

[0013] The sensor module is designed to capture a wide range of parameters. It preferably includes sensors for recording electrical parameters, such as current and voltage, as well as sensors for mechanical parameters, including vibrations, forces, and torques. These sensors are carefully selected and arranged to ensure comprehensive monitoring of the machine's critical functions and conditions. By integrating various sensor types, the module allows for a holistic view of the machine's operating status.

[0014] To ensure reliable and precise data acquisition, the sensor module can preferably be equipped with appropriate electronics and software. These include interfaces for data transmission, processors for preprocessing the data, and possibly storage media for temporary data storage. The acquired data is then forwarded, for example, via predefined protocols to the transmission module, which is responsible for the secure and efficient transmission of the data to the central storage module.

[0015] By strategically placing the sensor module on or near critical machine components, optimal data collection can be achieved, enabling accurate service life prediction. Advantageously, the sensor module is designed to be robust and resistant to the machine's operating conditions, ensuring long-term, reliable operation.

[0016] For the purposes of this patent application, a parameter is a quantifiable data type that describes an aspect of a machine's operating condition or performance, and whose measurement or estimation can provide information about the current condition assessment and expected service life of the machine. Parameters serve to identify, measure, and evaluate the quantities essential to the operation and integrity of the machine in order to be able to make prognostic analyses and decisions based on them.

[0017] The function of parameters within the machine life prediction system is to provide a data-based basis for analyzing and predicting the condition and performance of the machine. Parameters can reflect both physical and operational aspects, such as temperature, pressure, vibration levels, current and voltage values, and load factors. They thus reflect the actual operating conditions and loads to which the machine is exposed and serve as indicators of wear and tear and potential malfunctions.

[0018] The structure of a parameter can consist of its definition, the measurement or recording methods, and / or the evaluation criteria. The definition preferably includes a precise description of the aspect to be measured, its relevant physical unit, and the expected data range under normal operating conditions. The measurement or recording methods advantageously refer to the technical means and procedures used to determine the parameter value. This can include direct physical measurements using sensors or indirect estimations based on model calculations and algorithms. The evaluation criteria preferably specify how the recorded data is interpreted with regard to its impact on the service life and performance of the machine. Limit values ​​or patterns are preferably defined; if exceeded or detected, this indicates potential problems or maintenance requirements.

[0019] Preferably, parameters are selected and defined to provide a comprehensive and meaningful overview of the machine's condition. By continuously monitoring and analyzing these parameters, the system can provide valuable information for predicting a machine's expected service life, helping to optimize maintenance and extend its operating life.

[0020] For the purposes of this patent application, a transmission module is a dedicated component within the system for predicting the expected service life of a machine, whose primary function lies in the secure and efficient transmission of data acquired by sensors, and preferably also other relevant information, from the location of their acquisition to a central storage module or processing unit. The function of the transmission module is to collect the parameters acquired by the sensors of the sensor module, as well as preferably any additional information that may be provided by the input module, and to transmit this data to the central storage module and / or directly to the prediction module. Ensuring data security during transmission is of utmost importance to prevent manipulation or data loss.The module includes error detection and correction mechanisms that ensure that the transmitted data is correct and complete.

[0021] The transmission module preferably includes hardware components such as network interfaces or wireless communication modules that enable the sensors to communicate with the central storage module or other system components. Wireless data transmission technologies such as Wi-Fi, Bluetooth, or specific industrial communication standards are preferred, enabling flexible and scalable networking of system components. Furthermore, the module often includes software components or firmware that control transmission protocols, data encryption, authentication, and similar tasks to ensure secure data transmission.

[0022] Advantageously, the transmission module has a modular architecture that allows for easy integration into various use cases and adaptation to different security requirements. This also includes the ability to react to connection interruptions or failures, for example, by temporarily storing data and automatically resuming transmission once the connection is restored.

[0023] For the purposes of this patent application, a memory module is a component within a machine lifespan prediction system that is responsible for storing data collected from sensors and other sources. The memory module enables reliable, long-term storage of large amounts of data, enabling continuous analysis and evaluation of the collected information. The primary function of the memory module is to serve as a central repository for all data collected in the system. This includes data collected by the sensor module's sensors regarding machine performance and condition, such as temperature, vibration, electrical parameters, and other relevant measurements.In addition, the storage module can also be used to store configuration data, historical data and evaluation results, providing a comprehensive data basis for analysis and forecasting.

[0024] The storage module preferably includes physical storage media such as hard disk drives (HDDs), solid-state drives (SSDs), or special storage arrays specifically designed for the requirements of data management in industrial environments. These storage media are engineered to ensure the integrity and availability of the stored data, even under demanding operating conditions. Furthermore, depending on the specific system configuration, the storage module can be extended to cloud-based storage systems, which offer additional flexibility in terms of accessibility, scalability, and data backup.

[0025] On the software side, the storage module includes data management systems that enable effective organization, indexing, and rapid retrieval of stored data. These systems also advantageously support features such as data replication, automatic backups, and data encryption to increase data security against unauthorized access and data loss.

[0026] Advantageously, the storage module can be designed to support a hierarchical data structure that enables efficient data storage and retrieval. This may include the implementation of hot and cold storage solutions, where frequently accessed data (hot data) is kept on higher-performance storage media for quick access, while less frequently accessed data (cold data) is archived for the long term on more cost-effective media.

[0027] For the purposes of this patent application, an input module is a component of the machine life prediction system specifically designed to receive and process both manual and automated data inputs. This incoming data represents additional information essential to the accuracy and effectiveness of the life prediction, such as the machine's mechanical model, operating parameters, maintenance histories, and the like. The main function of the input module is to provide a user-friendly interface for entering data sets relevant to the system's prediction algorithms. It allows users to submit specific information about the machine structure, typical operating conditions, and intended usage scenarios, which the prediction module uses to improve the forecast accuracy.In addition, it is used to update the system database with new data points that come from external sources or are required due to changes in operating conditions or machine specifications.

[0028] The input module preferably comprises a combination of hardware and software components. On the hardware side, it can include various devices used for data entry, such as keyboards, touchscreens, or special input devices for industrial applications. These are supplemented by appropriate interfaces that enable connection to the overarching system and ensure smooth transmission of the entered data. On the software side, the input module advantageously includes application software or user interfaces specifically designed to facilitate the entry, validation, and preprocessing of user data. This preferably includes functions for error checking, formatting data according to system requirements, and supporting users with interactive help and instructions to ensure the quality and completeness of the entered information.Advantageously, the input module supports various data formats and input methods to ensure high compatibility and flexibility when dealing with different data sources and user preferences. It can also be equipped with data synchronization functions that enable seamless integration of manually entered data with the automatically collected data from the sensor module, creating a unified database for lifetime prediction.

[0029] For the purposes of this patent application, a mechanical model is an abstracted and preferably computer-based representation of the physical and functional properties of a machine, including, in particular, its design, operating behavior, and response to different load conditions. It is preferably used to simulate and predict the complex mechanical interactions within the machine and with its environment. The function of a mechanical model within the system for predicting the lifetime of a machine is to enable an in-depth analysis and understanding of the physical principles that govern the operation and wear mechanisms of the machine in question. The model provides valuable insights into mechanical resilience, potential weak points, and expected wear patterns under various operating conditions.This enables a more accurate and specific service life prediction tailored to the respective machine.

[0030] The construction of a mechanical model preferably includes a combination of geometric models, material models, and kinematic / dynamic modeling. Geometric models represent the physical shape and size of the machine and its components. They form the basis for evaluating stresses and deformations under load. Material models describe the material properties of the machine parts, including strength, hardness, elasticity, and plasticity, and are crucial for the analysis of material fatigue and fracture mechanics. Kinematic and dynamic modeling simulates the movement and interaction of the machine parts under operating loads, including the capture of equations of motion, inertial forces, and damping effects.

[0031] Advantageously, the mechanical model is implemented using powerful simulation software that utilizes numerical methods such as finite element analysis (FEA) or multibody simulation (MBS) to simulate the physical conditions under realistic or even extreme operating conditions. This software enables a visual representation of the results, such as stress fields or vibration modes, thus facilitating the identification of critical areas and the optimization of the machine design.

[0032] The model can further be compared and updated with real operating data collected by the sensor module to improve the accuracy of predictions and adapt the model to actual operating conditions.

[0033] For the purposes of this patent application, a first prediction module is a component of the system for predicting the expected service life of a machine, which is designed to provide an initial forecast of the remaining service life of the machine based on the collected and processed data. This module uses conventional analytical or algorithmic methods to derive conclusions about the condition and service life of the machine from the available data. The function of the first prediction module is to create a basis for the service life prediction by processing the parameters collected via the sensor module—such as temperature, pressure, vibration, and other service life-influencing variables—as well as the mechanical model of the machine stored in the input module.The module evaluates this data in the context of predefined algorithms or models based on physical laws, empirical data or statistical analyses to achieve an initial estimate of machine lifetime.

[0034] The structure of the first prediction module preferably includes an interplay of software components and mathematical models. The software components are responsible for data processing, the implementation of the calculation algorithms, and user interaction. They include interfaces for data transfer from the storage module and for outputting the forecast results to the output module. The mathematical models within the prediction module can range from relatively simple analytical equations to complex statistical or machine learning-based algorithms. This module is advantageously designed to be flexibly adapted to different machine types and operating conditions. This is achieved, for example, through the modular structure of the software code and the possibility of parameterizing the mathematical models.By adapting the models to specific machine characteristics and operating data, the accuracy of lifetime predictions can be significantly improved.

[0035] The first prediction module, together with the central storage module and the input module, forms the core of the prediction system. It enables the efficient and rapid creation of service life predictions, which serve as a starting point for preventive maintenance strategies and for decision-making in operational management. By providing an initial reliable forecast, the module helps operators better understand the condition of their machines and initiate appropriate measures to maximize machine service life and minimize unplanned downtime.

[0036] For the purposes of this patent application, a second prediction module is an enhanced component of the system for predicting the expected service life of a machine, specifically designed to provide a refined and more precise forecast of the machine's remaining service life through the use of AI-based technologies and algorithms. Unlike the first prediction module, which is based on traditional analytical and algorithmic methods, the second prediction module uses advanced data analysis techniques to gain deeper insights into the available data and identify complex patterns that are crucial for improved service life prediction.

[0037] The second prediction module performs detailed analysis of the collected machine data using machine learning and artificial intelligence methods. This takes into account not only directly measured parameters and the mechanical model of the machine, but also derived features and complex relationships between the data points. The module is capable of learning from historical and ongoing operational data to continuously improve the accuracy of service life predictions.

[0038] The structure of the second prediction module preferably comprises a combination of software components implementing AI algorithms and a data processing infrastructure capable of efficiently handling the computationally intensive operations. Preferred elements of these components can be selected from the following groups: Data processing units: These are responsible for preprocessing the input data, extracting and selecting relevant features, and normalizing the data to prepare it for AI processing. AI algorithms: The module advantageously includes a variety of machine learning algorithms, such as neural networks, decision trees, support vector machines, or ensemble methods, which can be selected and adapted depending on the application and data characteristics. Training and evaluation mechanisms: These enable learning from historical data and refinement of algorithms based on lifetime prediction results, including feedback mechanisms that allow iterative improvement of the models. User interfaces and visualization tools: These serve to present the results of the AI ​​analysis in an understandable and useful form and offer the user the opportunity to configure and adapt the model.

[0039] For the purposes of this patent application, a time series forecasting module is a specialized component within a system for predicting the expected service life of a machine. It is designed to forecast future values ​​of service life-influencing parameters based on historical data sequences and current measured values. This module specifically utilizes time series analysis and forecasting techniques to identify and evaluate patterns, trends, and seasonal fluctuations in the data, enabling a well-founded prediction of the development of the machine's service life. The function of the time series forecasting module is to model the predicted development of relevant machine parameters, such as temperature, vibration, or degree of wear, over time.By analyzing historical data sequences and incorporating current operating data, future machine conditions can be predicted with high accuracy. These forecasts are essential for planning and optimizing preventive maintenance measures, as they allow maintenance work to be scheduled before a potential failure or before reaching a critical wear level.

[0040] The structure of the time series forecasting module preferably includes the following elements: Data acquisition and preprocessing components: These components are responsible for collecting and processing data from the central storage module to prepare it for analysis. This includes eliminating outliers, normalizing the data, and segmenting it into trainable sequences. Modeling algorithms for time series analysis: The module advantageously integrates various advanced algorithms for time series forecasting, such as ARIMA (AutoRegressive Integrated Moving Average), Exponential Smoothing, and Long Short-Term Memory (LSTM) networks, which are suitable for different applications and data characteristics. The selection of the specific algorithm is based on the characteristics of the data and the forecasting goal. Learning and adaptation mechanisms: These mechanisms enable the module to learn from the results of previous forecasts and adapt the forecast models accordingly to improve forecast accuracy. This includes both the automatic adjustment of model parameters and the ability for the user to manually optimize settings. Output and visualization tools: After performing the time series analysis and forecast, the module provides tools for displaying the results. These tools allow for visualizing and interpreting the forecasted values ​​and trends in a user-friendly manner.

[0041] According to an advantageous embodiment of the invention, the sensor module can comprise sensors for detecting electrical parameters, in particular current and / or voltage, and / or mechanical parameters, in particular vibrations, forces, and torques. This versatility in detection provides a comprehensive data basis that improves the accuracy of service life predictions. Precise monitoring of parameters such as current, voltage, vibrations, forces, and torques allows early detection of signs of wear or defects, which promotes proactive maintenance measures and prevents potentially costly damage.

[0042] According to a further preferred development of the invention, the central storage module can also be configured as cloud storage. Configuring the central storage module as cloud storage offers several advantages, including scalable storage capacities, increased data security through redundancies, and improved accessibility of the stored data. This enables central analysis and evaluation of machine data across different locations, enabling, for example, cross-border maintenance strategies and service life predictions to be implemented. In addition, the cloud-based architecture facilitates the integration and updating of the prediction model, allowing the forecast capabilities to be dynamically adapted to changing operating conditions.

[0043] Furthermore, according to a similarly advantageous embodiment of the invention, the output module can be a PC, tablet computer, or smartphone, which can contribute to increasing the flexibility and user-friendliness of the system. By using these everyday devices as an interface, a broad user base can be reached without additional hardware investments. This also promotes the acceptance and integration of the technology into existing maintenance and operational processes, allows for timely monitoring and intervention by personnel, and supports decision-making through immediate access to relevant information.

[0044] According to a further particularly preferred embodiment of the invention, it can be provided that the system comprises a plurality of machines selected from the group of electrical machines, pumps, gears, etc.

[0045] Expanding the system to monitor multiple machines offers the advantage of universal applicability. The ability to integrate diverse machine types allows companies that use different types of machines in their operations to benefit from centralized lifespan prediction. This significantly reduces the complexity of machine management and optimizes maintenance planning. The universal applicability of the system thus contributes to a significant increase in efficiency in the monitoring and maintenance of machine fleets, which in turn increases operational reliability and reduces costs.

[0046] Furthermore, the invention can also be further developed such that the central storage module has an interface through which a user can load parameters influencing the service life of the machine into the central storage module. This enables users to dynamically adapt the system to changing operating conditions or new insights into factors influencing service life. The flexibility to expand or modify the data basis of the prediction model based on user input continuously improves forecast accuracy and relevance. Thus, this function supports the adaptability of the system to individual machine configurations and operating modes, enabling a more accurate and personalized service life forecast and ultimately maximizing operational efficiency and machine reliability.Historical data of a machine can also be loaded into the central storage module before its integration into the system according to the invention.

[0047] In a likewise preferred embodiment of the invention, it can also be provided that an AI-based prediction module for determining a second lifetime prediction within a system for predicting an expected lifetime of a machine according to claim, comprising • an input for receiving an input data set representing the parameters influencing the service life of the machine, • A time series forecasting module by means of which a prediction of the expected service life of the machine is determined from the parameters, and • an output for providing an output data set representing a lifetime prediction for a machine, which was predicted by the AI-based prediction module on the basis of the input data set.

[0048] The AI-based prediction module, equipped with an input data set input and a time series forecasting module, offers the significant advantage of adaptive and advanced analytics. By using AI, and specifically time series analysis, the module can detect complex patterns and trends in machine data that might be hidden from human observers. This enables a much more accurate prediction of machine lifespan, taking all relevant data points into account. By providing accurate predictions, companies can take preventive measures to prevent machine failures and thus minimize unexpected downtime costs and production losses.

[0049] It may also be advantageous to further develop the invention in such a way that an AI-based and computer-implemented method for determining a second lifetime prediction within a system for predicting an expected lifetime of a machine according to claim, comprising the following steps: Receiving an input data set representing the parameters influencing the service life of the machine, Carrying out a time series forecast by means of which an AI-based prediction of the expected service life of the machine is determined from the parameters, and Providing an output data set representing a second lifetime prediction for the machine

[0050] This computer-aided AI-based life prediction method enables standardized and automated analysis of large amounts of data, improving the accuracy and reproducibility of life predictions. Automating the process also increases efficiency by eliminating manual calculations and analyses, which can be time-consuming and error-prone. This optimizes maintenance and repair processes, extending machine life and reducing operating costs.

[0051] According to a further preferred embodiment of the subject matter of the invention, it can be provided that the time series forecast is based on a model selected from the group comprising ARIMA AutoRegressive Integrated Moving Average, Exponential Smoothing, and Long Short-Term Memory LSTM. By specifying that the time series forecast is based on models such as ARIMA, Exponential Smoothing, or LSTM, the adaptability and accuracy of the forecasting method are maximized. These models offer diverse approaches to analyzing and predicting time series data, allowing the system to respond flexibly to various patterns and anomalies in the machine's life cycle data. The ability to choose from a variety of proven models guarantees that the most suitable method is applied for the specific data and requirements of each machine, further increasing forecast accuracy.

[0052] Finally, the invention can also be advantageously implemented in such a way that a computer-implemented method for training an AI-based prediction model of an expected service life of a machine, in particular within a system for predicting an expected service life of a machine according to claim, comprising the following steps: • Receiving an input data set representing the parameters influencing the service life of the machine, • Training the AI-based prediction model with the input dataset.

[0053] The computer-implemented method for training the AI-based predictive model offers the significant advantage that the model continuously learns and adapts from real operating data. By receiving and analyzing the operating data, it ensures that the predictive model always incorporates the most up-to-date information and improves its forecast accuracy over time. This leads to continuous optimization of service life predictions, enabling companies to create more precise maintenance plans, thus minimizing operating costs and maximizing machine reliability.

[0054] The invention will be explained in more detail below with reference to figures without limiting the general inventive concept.

[0055] It shows: Fig. 1 an embodiment of a system for predicting an expected service life of a machine in a schematic block diagram, Fig. 2 a first lifetime prediction and a second lifetime prediction in a line diagram.

[0056] The Fig. 1 shows a system 1 for predicting the expected service life of a plurality of three machines 2a, 2b, 2c at a machine location 18. The machines 2a, 2b, 2c each have a sensor module 3a, 3b, 3c for continuously measuring parameters 4a, 4b, 4c that influence the service life of the machines 2a, 2b, 2c. The sensor module 3 can comprise various sensors for detecting electrical parameters, in particular current and / or voltage, and / or mechanical parameters, in particular vibrations, forces, and torques. The determined parameters 4a, 4b, 4c are transmitted in real time via a transmission module 6 to a central storage module 7 arranged in an IT infrastructure 17 and stored there.

[0057] The system 1 further comprises an input module 8 for receiving and storing data representing a mechanical model 9 of the machine 2, as well as a first prediction module 10, which is connected to the central storage module 7 and the input module 8 for determining a first service life prediction 11 of the machine 2 based on the determined parameters 4 and the mechanical model 9 of the machine 2. The system 1 further comprises an output module 12, via which the first service life prediction 11 of the first prediction module 10 can be visualized for a user. In the embodiment shown, the output module 12 is configured as a smartphone.

[0058] The central storage module 7, implemented as a cloud storage, is further connected to an AI-based second prediction module 13 for determining a second lifetime prediction 14 of the machine 2 based on the determined parameters 4. The second lifetime prediction 14 of the second prediction module 13 can also be visualized for a user via the output module 12. As a result, the user can then be provided with a lifetime forecast in the form of a line diagram, for example, as is also shown by way of example in the Fig. 2. The upper curve shows the first lifetime prediction 11 of the conventional prediction module 10, while the lower curve shows the second lifetime prediction 14 of the AI-based prediction module 13. It is clearly visible that the AI-based lifetime prediction 14 enables a more accurate forecast than the conventional lifetime prediction 11.

[0059] The central storage module 7 also has an interface 16 by means of which a user can load parameters 4 influencing the service life of the machine 2 into the central storage module 7.

[0060] The AI-based prediction module 13 for determining a second service life prediction 14 within the system 1 for predicting an expected service life of a machine 2 comprises an input for receiving an input data set which represents the parameters 4 influencing the service life of the machine 2, a time series prediction module 15 by means of which a prediction of the expected service life of the machine 2 is determined from the parameters 4, and an output for providing an output data set representing a service life prediction for a machine 2, which was predicted on the basis of the input data set by the AI-based prediction module 13.

[0061] To determine a second service life prediction 14, an input data set representing the parameters 4 influencing the service life of the machine 2 can first be received. Subsequently, a time series forecast can be performed, using which an AI-based prediction of the expected service life of the machine 2 is determined from the parameters 4. This prediction is then made available via a corresponding output data set. The time series forecast is preferably based on a model selected from the group comprising ARIMA AutoRegressive Integrated Moving Average, Exponential Smoothing, and Long Short-Term Memory LSTM.

[0062] To train the AI-based prediction model, an input data set can be received which represents the parameters 4 influencing the service life of the machine 2.

[0063] The invention is not limited to the embodiments illustrated in the figures. The above description is therefore not to be considered restrictive, but rather explanatory. The following claims are to be understood in such a way that a stated feature is present in at least one embodiment of the invention. This does not exclude the presence of further features. Where the claims and the above description define 'first' and 'second' features, this designation serves to distinguish between two similar features without establishing a priority. List of reference symbols 1 system 2 machines 3 Sensor module 4 parameters 5 Cloud environment 6 Transmission module 7 memory module 8 Input module 9 Model 10 Prediction module 11 Lifespan prediction 12 Output module 13 Prediction module 14 Lifespan prediction 15 Time Series Forecasting Module 16 Interface 17 IT infrastructure 18 Machine location

Claims

[1] System (1) for predicting an expected service life of a machine (2), wherein the machine (2) has a sensor module (3) for continuously measuring parameters (4) influencing the service life of the machine (2), wherein the parameters (4) can be transmitted via a transmission module (6) to a central storage module (7) and stored there, wherein the system (1) further has an input module (8) for receiving and storing data representing a mechanical model (9) of the machine (2), and a first prediction module (10) which is connected to the central storage module (7) and the input module (8) for determining a first service life prediction (11) of the machine (2) on the basis of the determined parameters (4) and the mechanical model (9) of the machine (2), wherein the system (1) further comprises an output module (12),via which the first lifetime prediction (11) of the first prediction module (10) can be visualized for a user, , characterized by that the central storage module (7) is further connected to an AI-based second prediction module (13) for determining a second service life prediction (14) of the machine (2) on the basis of the determined parameters (4), and the second service life prediction (14) of the second prediction module (13) can be visualized for a user via the output module (12). [2] System (1) according to claim 1, characterized by that the sensor module (3) comprises sensors for detecting electrical parameters, in particular current and / or voltage, and / or mechanical parameters, in particular vibrations, forces, moments. [3] System (1) according to one of the preceding claims, characterized by that the central storage module (7) is configured as cloud storage. [4] System (1) according to one of the preceding claims, characterized by that the output module (12) is a PC, tablet computer or a smartphone. [5] System (1) according to one of the preceding claims, characterized by that the system comprises a plurality of machines (2) selected from the group of electrical machines, pumps, gears, etc. [6] System (1) according to one of the preceding claims, characterized by that the central memory module (7) has an interface (16) by means of which a user can load parameters (4) influencing the service life of the machine (2) into the central memory module (7). [7] AI-based prediction module (13) for determining a second lifetime prediction (14) within a system (1) for predicting an expected lifetime of a machine (2) according to claim 1, comprising • an input for receiving an input data set representing the parameters (4) influencing the service life of the machine (2), • A time series forecast module (15) by means of which a forecast of the expected service life of the machine (2) is determined from the parameters (4), and • an output for providing an output data set representing a lifetime prediction for a machine (2), which was predicted on the basis of the input data set by the AI-based prediction module (13). [8] AI-based and computer-implemented method for determining a second lifetime prediction (14) within a system (1) for predicting an expected lifetime of a machine (2) according to claim 1, comprising the following steps: • Receiving an input data set representing the parameters (4) influencing the service life of the machine (2), • Carrying out a time series forecast by means of which an AI-based prediction of the expected service life of the machine (2) is determined from the parameters (4), and • Providing an output data set representing a second lifetime prediction for the machine (2) [9] Method according to claim 8, characterized by that the time series forecast is based on a model selected from the group comprising ARIMA (AutoRegressive Integrated Moving Average), Exponential Smoothing and Long Short-Term Memory (LSTM). [10] Computer-implemented method for training an AI-based prediction model of an expected service life of a machine (2), in particular within a system (1) for predicting an expected service life of a machine (2) according to claim 1, comprising the following steps: • Receiving an input data set representing the parameters (4) influencing the service life of the machine (2), • Training the AI-based prediction model with the input dataset.

Images