Fault mapping method

Abstract

The invention relates to a method for mapping faults in a maintenance management system, wherein the maintenance management system is used for a rail vehicle being examined. A language-based description of a detected fault in the rail vehicle is input into the maintenance management system in spoken form. The fault description is analyzed using a speech processing method which is automatically carried out, and input data is formed on the basis thereof. The input data is automatically mapped to known fault images of components of the rail vehicle using a database. It is logically and automatically checked, using the database, whether a fault image identified by means of the mapping matches the associated component. A check is also carried out in order to determine whether the mapping is logically meaningful in technical connection with other components of the rail vehicle. If the check result is positive, the maintenance management system displays the component which has been matched with the fault image in order to facilitate the repair of the component.

Description

[0001] 202501508

[0002] 1

[0003] Description

[0004] Error assignment procedure

[0005] The invention relates to a method for fault assignment in a maintenance management system used for rail vehicles.

[0006] Introduction and State of the Art

[0007] It is common practice to use a maintenance management system for the maintenance of rail vehicles. This system documents any faults detected in a serviced rail vehicle and, based on this documentation, initiates further maintenance steps.

[0008] For example, maintenance intervals for components of the rail vehicle are adjusted, future maintenance dates are planned, the procurement of replacement components is planned and initiated, etc.

[0009] Problems can easily arise when documenting errors in the maintenance management system or when assigning errors to components, which can prevent optimal maintenance management.

[0010] For example, maintenance personnel may negligently or due to time constraints attribute identified component faults to the wrong component, or the fault may be attributed to the rail vehicle instead of the component, resulting in incorrect fault and maintenance data being recorded in the maintenance management system. This, in turn, leads to inefficient resource management for the maintenance of the affected rail vehicle.

[0011] This problem will be illustrated and made more tangible using the following example: - A locomotive is taken to the workshop after the driver reports that the locomotive's pantograph is not raising correctly.

[0012] A maintenance technician begins troubleshooting and determines that the problem is not with the pantograph, but with a valve in the engine room that controls the compressed air supply for raising the pantograph. 202501508

[0013] 2

[0014] - When troubleshooting, the technician navigates through a tree structure of the locomotive, provided by the maintenance management system. This tree structure lists or contains main assemblies (e.g., the pantograph) and subassemblies (e.g., the valve installed in the engine room).

[0015] When correctly filling out a maintenance report in the maintenance management system

[0016] - the technician would describe or mark the sub-assembly "valve" as faulty, and

[0017] - Subsequently, a replacement would be procured for the defective valve and the defective valve in the engine room of the locomotive would be replaced.

[0018] If the maintenance report in the maintenance management system is filled out incorrectly, the technician would, for example, describe or mark the main assembly "current collector" as faulty.

[0019] Consequently, reliability statistics based on the maintenance management system would be distorted. A higher failure rate would be displayed for the pantograph, while a lower failure rate would be shown for the valve, contrary to the actual situation.

[0020] - For the valve, on the one hand, traceability of the actual fault would be lost, and on the other hand, future analyses of the reliability of similar pneumatic valves would be made more difficult or prevented.

[0021] - The planning of spare parts procurement and the strategy for maintaining larger vehicle fleets would be negatively affected simply because the actual fault was not correctly recorded.

[0022] To address this problem, maintenance personnel undergo repeated and extensive training in the use of the maintenance management system.

[0023] At the same time, maintenance personnel are trained to perform troubleshooting and fault documentation using standardized work instructions. This is intended to encourage maintenance technicians to write detailed fault reports and correctly select the affected components within the locomotive's hierarchical tree structure.

[0024] Nevertheless, the data entry process remains manual and prone to human error, especially when maintaining large fleets and under high workloads.

[0025] 3

[0026] Time pressure and limited resources often lead to errors, similar to the scenario described above.

[0027] In summary, this approach is inefficient and, in the long run, causes high inaccuracies in the data of the maintenance management system and in the subsequent statistical evaluation and maintenance planning.

[0028] Task

[0029] The object of the invention described below is to provide an improved method for fault assignment in a maintenance management system, which increases the accuracy in identifying faulty components and improves both data quality and the reliability of the maintenance management system.

[0030] This problem is solved by the features of claim 1. Advantageous further developments are specified in the dependent claims.

[0031] Description of the invention

[0032] The invention relates to a method for fault assignment in a maintenance management system, wherein the maintenance management system is used for a railway vehicle under investigation. A language-based description of a detected fault on the railway vehicle is entered into the maintenance management system in linguistic form. This fault description is analyzed using an automatically performed natural language processing method, and input data is generated based on this analysis.

[0033] Using a database, the input data is automatically matched to known fault patterns of rail vehicle components. The database then logically and automatically checks whether a fault pattern identified through the matching process corresponds to the assigned component. This check also verifies whether this match is logically meaningful in the technical context of other rail vehicle components.

[0034] Upon successful completion of the test, the maintenance management system displays the component that corresponds to the fault pattern, enabling its repair. 202501508

[0035] 4

[0036] In a preferred training method, a Natural Language Processing (NLP) method is used for language processing, which, based on a semantic context of the error description, recognizes a component name and / or an error symptom and translates it into input data.

[0037] In a preferred training course, at least one of the following content will be made available in the database:

[0038] - Fault patterns of components of the rail vehicle,

[0039] - Repair work assigned to each fault pattern,

[0040] - previously known reliability data of the components, and / or

[0041] - Functional dependencies between components of the rail vehicle.

[0042] The term "fault patterns" refers in particular to a description of the fault, a typical effect of the fault on the component, and typical accompanying circumstances of the fault (e.g., temperature development, noise development, etc.).

[0043] In a preferred advanced training method, the logical testing is carried out based on integrated machine learning algorithms or through the application of artificial intelligence, which is trained with a large number of real-world error cases to be able to calculate probabilities for a correct assignment of components based on the entered error description.

[0044] In a preferred further development, the algorithms take into account repairs carried out on components of the rail vehicle, operating conditions of the components that led to a component failure, and states of the component before a failure occurred.

[0045] In a preferred training course, if the component is successfully repaired, the database and algorithms are updated.

[0046] In a preferred training approach, the maintenance management system is part of a computer or implemented as a program within it, or it is part of a computer-based cloud application or runs on a processor, etc. 202501508

[0047] 5

[0048] Advantages:

[0049] The present invention increases the accuracy in identifying defective components.

[0050] The present invention improves the reliability of the maintenance management system.

[0051] The present invention performs a dynamic error assignment based on a combination of database matching, text analysis, and machine learning.

[0052] The present invention provides active verification and validation of entered error data and intelligent support for the technician in assigning errors to affected components.

[0053] The present invention detects discrepancies during input and provides the technician with real-time guidance on how to correct them.

[0054] The present invention increases the accuracy of the recorded data, improves the traceability of errors, and enhances the quality of maintenance management.

[0055] The present invention performs an intelligent comparison between human input and technical conditions.

[0056] The present invention uses historical data to verify coherence between fault description, selected components and repairs performed.

[0057] The present invention takes into account known dependencies between different assemblies on the one hand and their typical failure patterns on the other. This enables the present invention to understand technological relationships and to make alternative suggestions when an initial assignment of a failure to a component appears illogical.

[0058] Existing systems often require extensive manual checks to correct incorrect assignments later. The present invention automates this process, so202501508

[0059] 6

[0060] that not only the accuracy of the recorded data, but also the efficiency of maintenance is significantly increased.

[0061] The present invention, which operates in real time, makes it possible to provide a technician with feedback and suggestions during data entry. This makes it possible to avoid errors before they are saved.

[0062] Character description:

[0063] The invention is explained in more detail below with the aid of a drawing. The single figure FIG shows an exemplary embodiment of the invention.

[0064] A technician is examining a rail vehicle for a reported fault.

[0065] In a first step S1, the fault-finding technician enters a language-based fault description into a maintenance system or a maintenance management system.

[0066] For example, this includes the error description "Pantrone does not lift".

[0067] For this, the technician uses a human-machine interface, such as a microphone or a keyboard.

[0068] The maintenance system or maintenance management system is preferably part of a computer, or part of a computer-based cloud application, or runs on a processor, etc.

[0069] In a second step S2, the error description is analyzed using a language processing method and input data is generated based on this.

[0070] For example, the maintenance system uses natural language processing (NLP) for language processing. This NLP is trained to recognize a component name (e.g., "pantograph") and / or a fault symptom ("no lift") based on the semantic context of the fault description, which are then used as input data.

[0071] 7

[0072] In a third step S3, the maintenance system automatically assigns the input data to known fault patterns of components of the rail vehicle under investigation.

[0073] For example, the input data “current collector” and “no lift” are assigned to a previously known error pattern “defective pneumatic valve in the engine room”.

[0074] The error patterns are, for example, stored in a database of the maintenance system or are requested from a database by the maintenance system for this purpose.

[0075] In the database

[0076] Fault patterns of the components of the rail vehicle are preferably stored in the form of a history,

[0077] Are repair work associated with each fault pattern stored, are known reliability data associated with each component stored, and / or

[0078] Functional dependencies between the components of the rail vehicle are stored.

[0079] Using the database, a logical check is performed to determine whether the fault pattern assigned via the input data matches a component or not. Simultaneously, it is checked whether the assignment is also technically meaningful in relation to other components of the rail vehicle.

[0080] This ensures that an error pattern actually corresponds to the behavior of an associated component, or whether the error pattern points to another component that is indirectly or directly related to it.

[0081] This step is supported by integrated machine learning algorithms or by a corresponding application of artificial intelligence (AI). The algorithms are trained on a multitude of real-world error scenarios and are thus able to calculate probabilities for the correct assignment of components based on the input error description.

[0082] The algorithms take into account repairs carried out on components, operating conditions of the components that have led to a component failure in the past.

[0083] 8

[0084] led to, as well as the state of the component prior to the occurrence of a fault in the component - for example, a rapidly rising component temperature.

[0085] In a fourth step S4, if a logical test (see step S3) is passed, the fault-finding technician is informed of the component's assignment to the fault pattern, which was made in the third step S3, in order to enable the technician to repair the component.

[0086] Referring to the example, the service technician is informed of this assignment: "defective pneumatic valve in the engine room".

[0087] In a fifth step, S5, the repair is carried out. The technician then confirms either a successful repair of the component or an unsuccessful repair.

[0088] In a sixth step, S6, this confirmation is evaluated by the maintenance system. This allows the database to be updated with respect to the component. This, in turn, enables the artificial intelligence to improve its underlying algorithms and thus enhance its machine learning.

Claims

202501508 9 Patent claims 1. Procedure for fault assignment in a maintenance management system, - where a maintenance management system is used for a rail vehicle under investigation, - where a language-based description of a detected fault on the rail vehicle is entered into the maintenance management system in linguistic form, - where the fault description is analyzed using an automatically performed speech processing procedure and input data is generated based on this, - in which an automatic assignment of input data to previously known fault patterns of components of the rail vehicle is carried out using a database, - where, using the database, it is logically and automatically checked whether a fault pattern recognized through the assignment matches the assigned component, and where it is checked whether this assignment is logically meaningful in the technical context with other components of the rail vehicle, - where, upon successful completion of the test, the component is displayed that is assigned according to the fault pattern in order to enable a repair of the component.

2. The method according to claim 1, wherein a natural language processing method referred to as NLP is used as the language processing method, with which a component designation and / or a fault symptom is recognized and converted into input data based on a semantic context of the fault description.

3. The method of claim 1, wherein at least one of the following contents is made available in the database: - Fault patterns of components of the rail vehicle, - Repair work assigned to each fault pattern, - previously known reliability data of the components, and / or - Functional dependencies between components of the rail vehicle.

4. Method according to claim 1, wherein the logical testing is based on integrated machine learning algorithms or by an application of artificial intelligence. 10 is carried out, which are trained with a large number of real-world error cases in order to be able to calculate probabilities for a correct assignment of components based on the entered error description.

5. The method of claim 4, wherein the algorithms take into account repairs made to components, operating conditions of the components that led to a component failure, and states of the component prior to the occurrence of a component failure.

6. Method according to one of the preceding claims, wherein, upon successful repair of the component, an update of the database and the algorithms is performed.

7. Maintenance management system comprising means for carrying out the method according to any one of claims 1 to 6.

Images