Predictive fault detection system for a mobile work machine

Abstract

System (1) for predictive diagnosis of fault conditions of components (5) of the powertrain of a fleet (3) of electrically powered mobile working machines (4), wherein the system comprises: at least one working machine (4) with a communication module for sending operating data (BG); and a central unit (2) spatially separate from the working machine (4), which is designed to receive the operating data (BG) from the working machine (4) via a communication network, characterized by a data processing unit implemented in the central unit (2) with: a) an anomaly detection module configured to detect deviations from normal behavior in the operating data (BG) received from the working machine (4) that indicate an impending fault; and b) at least one hybrid fault model (24) configured to analyze the relevant operational data (OD) after detection of a deviation by the anomaly detection module, wherein the hybrid fault model (24) combines physical domain knowledge about the component (5) and data-driven models to specify a fault type and determine a fault severity.

Description

TECHNICAL AREA

[0001] The invention relates to the technical field of monitoring and diagnostics of electrical systems in vehicles. In particular, the invention relates to a system for predictive fault detection for powertrain components of one or more electrically powered mobile work machines, such as electric excavators, electric wheel loaders, or electric bulldozers. STATE OF THE ART

[0002] Electrically powered mobile machinery represents an increasingly important alternative to internal combustion engine-powered machines, particularly with regard to emission reduction and noise abatement. These machines are frequently used in demanding and difficult-to-access environments such as construction sites, mines, or forestry operations. The reliability and availability of the electric powertrain are therefore of crucial economic importance. An unexpected failure of a critical component can leave the machine stranded in difficult terrain, resulting in significant downtime, high recovery and repair costs, and delays in the overall work process.

[0003] The electric drive train of such work machines consists of several key components, including an electrical energy storage device such as a high-voltage battery, one or more power electronics units (inverters) for controlling the motors, and one or more electric machines (drive motors). Each of these components is susceptible to wear and potential defects.

[0004] Monitoring systems integrated into the control units of the machine are known in the prior art. One example is the Battery Control Unit (BCU) for monitoring the high-voltage battery. These systems typically monitor parameters such as the battery's state of health (SoH). However, an error message is usually only generated when a fixed, predefined threshold (e.g., an SoH of 80%) is exceeded or fallen below. By this time, the component is often already significantly and irreversibly damaged. The fault is already far advanced, and imminent total failure is likely. This reactive approach offers no predictive warning and leaves the operator little time to safely move the machine out of the work area or to proactively plan maintenance.

[0005] Another approach to increasing reliability is the redundant design of components. For example, the entire powertrain could be duplicated. However, this leads to significant disadvantages in terms of installation space, weight, and above all, cost, since the high-voltage battery in particular is one of the most expensive components of the vehicle.

[0006] Modern work machines increasingly feature connectivity solutions that enable data transmission to an external central unit, such as a cloud. This infrastructure is currently used primarily for non-time-critical tasks like general state of health (SoH) determination or fleet management functions. However, a system that specifically utilizes this connectivity for in-depth, predictive fault diagnosis that goes beyond the capabilities of on-board control units is lacking. SUMMARY

[0007] The object of the present invention is to provide a system and a method which overcome the disadvantages of the prior art and enable early, reliable and predictive diagnosis of faults in components of the drive train of electric working machines.

[0008] This task is accomplished by a system having the features of claim 1. Advantageous embodiments and further developments of the invention are specified in the dependent claims.

[0009] The core of the invention lies in the creation of a system in which a central computing unit—a so-called central processing unit or cloud—located spatially separate from the machines, collects and analyzes the operational data of an entire fleet of machines. This central processing unit employs a software-based, hybrid fault model that combines physical domain knowledge (e.g., about the aging behavior of a battery) with powerful, data-driven models (e.g., AI models).

[0010] This combination makes it possible to identify complex, multivariable relationships in operational data that indicate an impending fault, long before individual parameters reach critical thresholds. By centrally aggregating data from the entire fleet, the system can learn from real-world field data, continuously refine its models, and thus achieve a level of diagnostic accuracy that would be impossible with in-vehicle systems.

[0011] A first step in the analysis within the central unit is anomaly detection, which identifies unusual patterns or trends in the operational data. If such an anomaly is detected, a more in-depth analysis is triggered by the specific hybrid fault model. This model is capable of clearly specifying the fault, assessing its severity, and predicting its impact on the remaining service life of the component and thus the remaining operating time of the machine. This enables the fleet operator to proactively plan maintenance, minimize downtime, and maximize operational reliability. BRIEF DESCRIPTION OF THE DRAWING

[0012] The invention is explained in more detail below with reference to the accompanying drawings, which show an advantageous embodiment of the invention. Fig. Figure 1 shows a schematic representation of a system for the predictive determination of fault states of drive train components of a fleet of mobile working machines according to an embodiment of the invention. DETAILED DESCRIPTION

[0013] The present invention is described below with reference to certain embodiments. However, the present invention is not limited to the specific embodiments described in the following detailed description; rather, the described embodiments merely illustrate some aspects of the present invention, the scope of which is defined by the claims.

[0014] Further modifications and variations of the present invention are obvious to a person skilled in the art. The present description therefore encompasses all modifications and / or variations of the present invention whose scope of protection is defined by the claims.

[0015] Fig. Figure 1 shows a system 1 for the predictive detection of fault conditions in components of a large number of electrically powered mobile work machines 4. The system 1 comprises a central unit 2, which is configured to receive and process data from a fleet 3 of work machines 4. In this embodiment, the mobile work machines 4 are electric excavators (e-excavators). The central unit 2, preferably implemented as a cloud-based computing unit, enables early fault detection, determination of the fault severity, and calculation of the remaining operating time (ROT) for each individual work machine 4 in the fleet 3. The aim is to perform the earliest possible and most precise fault diagnosis of the drive train components and other relevant electrical components based on the operating data (BG) of each work machine 4.

[0016] One of the work machines 4 of fleet 3 is in Fig. 1 is shown in more detail as an example and connected to a drive system 5. The drive system 5 comprises the essential components for the electric drive and the energy supply of the working machine 4. This drive system 5 includes a high-voltage battery as a rechargeable electrical energy storage device, a first inverter, and a first electric drive motor.

[0017] In a preferred embodiment, such as fully electric excavators where both propulsion (driving) and work processes (e.g., digging, lifting) are purely electric, the machine 4 features an extended drive train. This includes, in addition to the first inverter and the first drive motor (e.g., for the travel drive), a second inverter and a second drive motor (e.g., for the working hydraulics or another drive train). The high-voltage battery serves as the central energy storage unit and can be partitioned for the different operating modes. The presented diagnostic method is capable of monitoring and diagnosing all of the aforementioned components—that is, both inverters and both drive motors. Thus, faults in components of the electric travel drive, as well as in electrical components relevant to the work process, can be diagnosed.

[0018] Within the drive system 5 is a central vehicle-side control unit. This control unit is connected to a communication module capable of wirelessly transmitting operating data (OD) between the respective work machine 4 and the central unit 2. The vehicle-side control unit is bidirectionally connected to the subordinate control units of the individual components. These include a battery control unit (BCU), a control unit for the first inverter, and a control unit for the second inverter.

[0019] Each component is equipped with sensor units that include one or more sensors for recording operating parameters. For example, sensor units on the drive motors continuously record operating parameters such as speed, torque, or temperature and transmit this data to the respective inverter control unit.

[0020] The inverter control units also record their own operating parameters (e.g., current, voltage, temperature of the power electronics). Similarly, the BCU uses sensors on the high-voltage battery to record battery-specific operating parameters. These include, in particular, the current battery current, voltage, temperature, and state of charge (SoC), which can advantageously be recorded at different levels (pack, module, and / or cell level).

[0021] All recorded operating parameters are collected by the subordinate control units (BCU, inverter controllers) and transmitted to the central vehicle-side control unit. This unit then forwards the operating data (BG) to the central unit 2. The transmission can occur at regular intervals, for example, in blocks at intervals of a few minutes to several hours, and in uncompressed and / or compressed form. For the described method, it is advantageous that the working machines 4 transmit the characteristic operating parameters during all relevant operating states, i.e., during driving, during the work process, and also during charging.

[0022] The central processing unit 2 has a data processing unit in which various models and algorithms are implemented. In a first processing path, the received operating data BG is fed to an aging state model 21. For example, during a charging process of the high-voltage battery, the amount of charge supplied can be recorded and transmitted to the aging state model 21 along with the corresponding voltage range and battery temperature. This model determines the current aging state (state of health, SoH) of the respective component from the operating data; in the case of the battery, for example, the battery's SoH value.

[0023] The aging state model 21 is followed by an aging prediction model 22. This model 22 predicts the future aging process and determines a predicted remaining service life of the component. In the event of a detected fault, the remaining operating time 23 of the machine 4 is determined, taking the influence of the fault into account.

[0024] In parallel, the operating data BG is fed into a module for predictive diagnostics 24. This module 24 is the core of predictive fault detection and typically includes anomaly detection, a hybrid fault model, and a unit for determining the fault severity and the impact of the fault on the aging condition.

[0025] As an example of predictive diagnostics, consider the increase in a battery's internal resistance. This value is an indicator of battery aging and is associated with a rise in temperature. However, the battery temperature is influenced by many factors (e.g., load current, ambient temperature, state of charge). A gradual, impermissible change in internal resistance cannot be directly detected by sensors. Conventional systems only trigger a warning when a critical temperature threshold is exceeded. By this time, however, the fault is already well advanced. The method according to the invention makes it possible to detect such a fault much earlier by monitoring the fault parameter—in this case, the battery temperature—for an atypical trend, taking all relevant dependencies into account.Such complex, multivariate analyses of large datasets are advantageously handled by the use of AI models (data-driven models), especially in the form of hybrid models.

[0026] The predictive diagnostics in Module 24 work as follows: The operational data (BG) continuously received from Fleet 3 serve as input for an anomaly detection model. This data-driven model is trained to distinguish normal behavior from atypical behavior. It checks the plausibility of the data, including its correlation with each other, and identifies atypical trends or patterns that indicate an impending fault.

[0027] When an anomaly is detected, the relevant operating data and fault parameters are forwarded to a specific fault model responsible for that type of fault. A library of fault models exists, with each model describing a different fault and relying on different fault parameters as inputs (e.g., temperature and current for Fault_1; voltage and charge for Fault_2). These fault models are preferably designed as hybrid models that combine physical knowledge with data-driven approaches. They can uniquely specify the fault, set a corresponding fault suspicion bit, and determine its effects on the aging state (e.g., the state of health curve) as well as the severity of the fault.

[0028] To determine the severity of a fault, the fault model analyzes the historical development of the fault parameters and checks whether predefined threshold values ​​(SW 1...3) have been exceeded or fallen below at specific operating points. Furthermore, the model predicts the future development of the fault parameters. This prediction serves as input for predefined characteristic curves or maps that describe the impact of a fault on the aging state (e.g., the state of health value).

[0029] These characteristic maps are created through extensive testing on test benches before market launch (SOP - Start of Production). During these tests, faults are deliberately introduced into the components, their effects are measured, and the signal waveforms of relevant fault parameters are recorded. The effects at minimum and maximum fault levels are captured to define the limits of the characteristic maps. A key advantage of System 1 is that these characteristic maps do not need to be fully available before SOP. They can be dynamically supplemented and refined using data from fleet vehicles 3 where the corresponding fault occurs in real-world operation. Operating-point-dependent data points that were not recorded in the laboratory are thus supplemented with field data.

[0030] Another condition that can trigger an anomaly is an atypically sharp decline in the state of health (SoH) value, which is calculated in module 21 of the central unit 2. If the anomaly detection system recognizes that the decrease in the SoH value exceeds a threshold – either compared to the previous value of the same vehicle (SoH_vehicle(ti-1) - SoH_vehicle(ti) > SW) or compared to the fleet average (SoH_vehicle(ti) - SoH_fleetvehicles(ti) > SW) – the diagnostic process in module 24 is also initiated.

[0031] Under normal operating conditions, the SoH value is determined at longer intervals (e.g., once a week). However, if a suspected fault is reported (suspected fault bit is set), the calculation of the SoH value is intensified and performed at every possible opportunity to detect accelerated degradation and a resulting decrease in the remaining operating time (SoH) at an early stage.

[0032] The results of the predictive diagnostics, in particular the type of fault, the severity of the fault, and the remaining operating time 23, are processed in the central unit 2 and can be displayed, for example, via a visualization device to the fleet manager or the driver / operator of the work machine 4. The severity of the fault can be visualized, for example, by a traffic light system (green, yellow, red) to indicate the urgency of maintenance (e.g., "visit workshop").

[0033] Although the procedure is described here using the example of a battery and its state of health (SoH) value, it is universally applicable to other components and their specific health indicators. For example, in an electric drive motor, rotor vibrations or noise can serve as parameters to be monitored, indicating incipient bearing damage. The core of the procedure, the predictive diagnosis summarized in Module 24, remains structurally the same and can be applied to the monitoring of multiple components simultaneously.

[0034] While the present invention has been described with reference to the embodiments described above, it is clear to the person skilled in the art that it is possible to implement various modifications, variations and improvements of the present invention in light of the teaching described above and within the scope of the attached claims without deviating from the scope of protection of the invention.

[0035] Furthermore, the areas in which experts are likely to be knowledgeable have not been described here in order to avoid unnecessarily obscuring the described invention.

[0036] Accordingly, the invention should not be limited by the specific illustrative embodiments, but only by the scope of protection of the attached claims.

Images