A method and system for failure prediction using oil analysis results in construction machines

Abstract

The invention relates to a malfunction prediction system and a method for construction machines enabling prediction of faults based on the oil analysis values obtained from construction machines, and thus enabling the identification of maintenance needs for the machine.

Description

[0001] A METHOD AND SYSTEM FOR FAILURE PREDICTION USING OIL ANALYSIS RESULTS IN CONSTRUCTION MACHINES

[0002] Technical Field

[0003] The invention relates to a method and system for predicting faults in construction equipment by utilizing oil analysis data obtained from these machines, enabling early fault detection and identification of maintenance needs for the vehicles.

[0004] More specifically, the present invention pertains to a system that predicts when a construction machine will malfunction within a certain operating time by utilizing oil analysis data from the vehicles. It involves the early detection of potential malfunctions through machine learning methods based on machine oil analysis values. The predicted data is integrated into decision-making mechanisms within business processes, and the information obtained through data visualization technologies is reported and made traceable through summary data.

[0005] State of The Art

[0006] Advancements in technology contribute to the continuous development and expanding applications of machines. These machines undergo regular maintenance at specified intervals, ensuring necessary checks and procedures to prevent and address potential malfunctions. Periodic maintenance not only extends the lifespan of machines but also helps proactively avoid and resolve issues. Furthermore, these scheduled maintenance activities contribute to the efficient operation of the machines, ensuring that any arising problems are promptly addressed for optimal performance.

[0007] Machines operating seamlessly is crucial in terms of time, labor, and cost efficiency. Particularly, malfunctions in construction machinery lead to downtime, rendering tasks unachievable and causing disruptions in the workflow. Therefore, the rapid detection of malfunctions in construction machinery is essential. Identifying the faulty component and promptly delivering the required spare part will minimize the repair time. This, in turn, prevents business and financial losses caused by machinery breakdowns. Industry 4.0 has brought personalized services to the forefront in the field of machinery. One of these services is the practice of predicting machine failures in advance to prevent disruptions. Based on the results of these personalized services, technical teams can shift their focus from “fixing what's broken” to their core responsibility of “preventive maintenance” aiming to prevent breakdowns. However, the primary function of eliminating the fundamental sources of malfunctions is not to change the structure of maintenance teams but to significantly reduce maintenance costs and extend the life of machines and components.

[0008] Introduced in the 1980s as a cornerstone of advanced maintenance technology, Proactive Maintenance (Preventive Maintenance) is fundamentally based on oil maintenance. Studies indicate that 42% of mechanical failures stem from oil contamination and lubrication-related issues. Hence, conscious oil usage and preserving oil quality are crucial factors. Oil analyses play a vital role in achieving this, providing an essential component of a preventive maintenance program used to enhance the lifespan and efficiency of construction machinery, while preventing maintenance expenses and downtime. Oil analyses allow testing all used oils, including engine, transmission, differential, gear, and hydraulic system oils. These tests are developed not only to assess the condition of the oil but more importantly to evaluate the state of construction machinery and engines. The success of a preventive maintenance program is measured by the financial gains it brings to the users.

[0009] The application of oil analysis is not only cost-effective, quick, and easy but can also result in significant savings. When the practice of oil analysis is combined with the high- performance characteristics of the original oil and filters used, it forms the foundation of the best possible maintenance program for machines. In construction machinery, periodic machine oil analysis results are utilized to identify where the malfunction may be occurring within the machine. The particle data related to the oil contained in construction machinery is obtained through samples taken with these oil analyses and their subsequent laboratory results. Following the analysis, the decision to continue using the oil is made based on the identified qualities of the oil.

[0010] Existing methods typically rely on regular measurements of machine oil values using monitoring devices to determine whether there is a problem or not. However, the current solutions do not allow for early fault detection based on machine learning according to oil analysis values during routine maintenance intervals, prediction of when the machine will fail in terms of working hours, and identification of maintenance needs for construction machinery.

[0011] This situation necessitates the development of a method that, through the determination of when the machine will fail based on the values of machine oil analyses, addresses the need for using these data in the decision-making processes related to the nature of the fault. Additionally, the method ensures that the machine is traceable based on these data.

[0012] The patent document CN101832902B relates to a method for equipment fault diagnosis, particularly focusing on an oil analysis method for diagnosing equipment faults. The fault diagnosis technology based on oil analysis is specifically tailored for large hydraulic equipment. The oil analysis technology can identify the condition of the lubricant, the performance changes in the working environment of the analytical equipment or machine, and the presence of wear particles. This allows obtaining information about the lubrication and wear condition of the equipment, the operating mode of the evaluation device, and predicting and determining the exact cause, type, and location of the fault. In the invention, it is not possible to predict when the fault will occur based on the data obtained through oil analysis methods, even with the help of historical data through machine learning.

[0013] The patent document LIS2021088487 relates to a method and system for fault prediction using the analysis of oil or other lubricants. Raw data about the characteristics of numerous particles filtered from a liquid sample are utilized to categorize each particle into one of multiple categories. Each category is defined by one or more chemical compositions, sizes, and morphologies. The physical properties of particles in each category are qualified to obtain a set of categorized data. The categorized data is then compared with historical data, and the results of the comparison are evaluated to form a prediction of any fault or malfunction mechanism. Although the patent document utilizes oil analysis for fault prediction, it does not provide a prediction regarding when the machine will malfunction in terms of working hours.

[0014] The patent document CN113218903 relates to a method for oil fluid detection analysis and equipment evaluation fault prediction, specifically within the intersection of microfluidics and big data artificial intelligence technology. The invention describes a microfluidic and artificial intelligence-based fault prediction system for oil analysis equipment, including an oil sample input structure, grinding powder feature analysis chip, a first data collection unit, a petroleum physicochemical analysis chip, and a second data collection unit. The invention utilizes postanalysis cloud computing for data acquisition, and prediction is made through machine learning based on this data. It's important to note that the information is not stored in a single database and does not disclose when the machine will malfunction in terms of working hours.

[0015] Consequently, the emergence of a solution devoid of the disadvantages mentioned above has become necessary.

[0016] Objectives and Brief Description of the Invention

[0017] The purpose of the invention is to predict the presence and timing of malfunctions in machinery or vehicles, particularly those with internal combustion engines, such as construction equipment, before they occur.

[0018] Another objective of the invention is to take actions to address the technician and spare parts needs based on the identified nature of the malfunction.

[0019] Another purpose of the invention is to enhance the lifespan and efficiency of the construction machine by planning its operational sequence.

[0020] Another aim of the invention is to reduce maintenance and repair costs, preventing work and time losses.

[0021] Another objective of the invention is to minimize the downtime of machines when they are inoperable, preventing customers from experiencing time losses in service visits.

[0022] The invention, in order to achieve the aforementioned objectives, is a fault prediction system for construction machinery that utilizes oil analysis values / results to detect faults. The system comprises the following elements:

[0023] - at least one CRM system containing machine / model information and fault records,

[0024] - at least one computing device into which oil analysis values / results obtained from the construction machinery are entered,

[0025] - at least one database that logically stores and manages machine / model information, fault records, and oil analysis values / results, - at least one server that formats data coming from the database for processing,

[0026] - at least one prediction component within the server that enables the operation of the prediction algorithm.

[0027] The system comprises a prediction component that processes the information in the mentioned database using a decision tree model in the Python programming language

[0028] The invention is also a fault prediction method for construction machinery, using oil analysis values for fault detection, and it comprises the following steps:

[0029] - entering oil analysis values / results of an oil sample taken from a construction machinery into a computing device and transferring them to a database,

[0030] - sending machine / model information and fault records of the construction machinery from which the oil sample was taken, through a CRM system to the database,

[0031] - analyzing the data received from the computing device and the CRM system in the database, converting it into a single format, and transferring it to a server,

[0032] - preparing the data coming from the database in the server for machine learning and creating the prediction model,

[0033] - running an algorithm by a prediction component within the server to detect the presence of a fault and to detect the component in which the fault is present,

[0034] - sharing prediction result with responsible business units through the server after detecting the presence and location of the fault,

[0035] - recording the prediction result in the fault records within the CRM system after detecting the presence and location of the fault.

[0036] Brief Description of the Figures

[0037] Figure 1 shows the system components of the system subject to the invention and the relationship between them.

[0038] Figure 2 shows the steps related to the method subject to the invention.

[0039] Reference Numbers

[0040] 10. Fault prediction system 11 . CRM system

[0041] 11.1 Machine / model information

[0042] 11 .2 Fault records

[0043] 12. Computing device

[0044] 12.1 Oil analysis values / results

[0045] 13. Database

[0046] 14. Server

[0047] 14.1 Prediction component

[0048] 15. Prediction result

[0049] 100. Fault prediction method

[0050] 101 . Entering the oil analysis values / results of the oil sample taken from the construction machinery into the computing device and transferring them to the database

[0051] 102. Sending the machine / model information and fault records of the construction machinery from which the oil sample was taken, through the CRM system to the database

[0052] 103. Analyzing the data received from the computing device and CRM systemin the database, converting it into a single format, and transferring it to the server

[0053] 104. Preparing the data coming from the database in the server for machine learning and creating the prediction model

[0054] 105. Running the algorithm by the prediction component within the server to detect the presence of a fault and in which component it is located

[0055] 106. Sharing the prediction result with the responsible business units through the server after detecting the presence of a fault and in which component it is located 107. Recording the prediction result in the fault records within the CRM system after detecting the presence of a fault and in which component it is located.

[0056] Detailed Description of the Invention

[0057] The invention pertains to a method (100) and system (10) for making a fault prediction using oil analysis values / results (12.1) obtained from construction machines, enabling the prediction of when the construction machine will malfunction within how many working hours and early detection of faults related to these machines, and defining the maintenance needs of the construction machine.

[0058] The existing invention is related to a system (10) allowing the prediction of when the machine will malfunction within a certain number of working hours through machine learning methods by using oil analysis values / results (12.1) related to construction machines. It involves integrating prediction data into decision mechanisms in business processes, reporting the obtained information using data visualization technologies, and making the information traceable through summary data.

[0059] In the present invention, oil samples are collected from construction machines at specific intervals. These oil samples are then analyzed in laboratories based on the systems they operate in. During the analysis, the oil samples may contain 26 types of particles related to the systems they operate in (e.g., K, CU, Cr, Na, Ca, Sn, Cu, Zn, Fe, Water). The type and concentration of these particles provide clues about which component of the systems is experiencing a malfunction. For example, if the "Cu" value exceeds a threshold, it can be anticipated that values above this threshold are due to wear in the engine.

[0060] Oil analysis can detect contaminants such as wear particles, oxidation, nitration by-products, sulfur by-products (acids), as well as the presence of water, fuel, and antifreeze mixtures. Through oil condition analysis, it is possible to assess how much the oil has deteriorated during use and determine its compliance with specifications over that period. In oil analysis, the detection of an increase in the concentration of one or more metals can also indicate the component that may be causing the increase. For example, a sudden rise in copper and iron values in a hydraulic oil sample may indicate wear in the hydraulic pump due to oil degradation or contamination. In the invention, the obtained oil analysis values / results (12.1) are transferred to the database (13) through a computing system (12), along with machine / model information (11.1) and fault record (11.2) information within the CRM system (11). A data manipulation process is performed to modify the data for easier readability and organization in the transferred database (13). Utilizing machine learning methods, the accumulated data up to the past oil analysis changes is analyzed based on the oil analysis values / results (12.1) and historical fault records (11.2), resulting in predictions according to the mentioned data intervals. This enables the anticipation of when the machine will malfunction based on the analysis of past fault records, estimating how many working hours remain until a breakdown.

[0061] The aforementioned system (10) used in the described fault prediction method comprises;

[0062] - at least one CRM system (11) comprising machine / model information (11.1) and fault records (11.2),

[0063] - at least one computing device (12) into which oil analysis values / results (12.1) obtained from the construction machinery are entered,

[0064] - at least one database (13) that logically stores and manages machine / model information (11.1), fault records (11.2), and oil analysis values / results (12.1),

[0065] - at least one server (14) for formatting and processing data coming from the database (13),

[0066] - at least one prediction component (14.1) within the server (14) that enables the operation of the prediction algorithm,

[0067] - prediction result (15) that provided by running and applying the algorithm in the prediction component (14.1).

[0068] The CRM system (11) connects and harmonizes large amounts of data from various sources, serving as a digital storage system. Additionally, after the detection of the fault and identification of the affected component, the fault detection information obtained from the prediction result (15) is sent to the CRM system (11) to be recorded.

[0069] The fault records (11.2) contain information related to the faults encountered during the operation of the construction machine.

[0070] The computing device (12) refers to computer hardware and an internet platform for processing the values / results after the oil analysis (12.1). The database (13) organizes, updates, manages, and makes data available in an orderly manner. It also archives the data logically for storage purposes.

[0071] The server (14) undergoes preprocessing steps by the prediction component (14.1) obtained from the database (13) to make the data processable. These steps include cleaning and normalizing the data. Additionally, it makes the necessary adjustments to store the data correctly.

[0072] The prediction component (14.1) selects a model from predefined models and trains this model with the data. Due to the discrete nature of the data, it uses the decision tree algorithm within machine learning algorithms. It creates a prediction model to anticipate potential issues and predict when the construction machine will malfunction after a certain number of working hours.

[0073] The method (100) for predicting faults using oil analysis results in construction machinery comprises the following steps;

[0074] - entering the oil analysis values / results (12.1) of an oil sample taken from a construction machinery into a computing device (12) and transferring them to the database (13) (101),

[0075] - sending machine / model information (11.1) and fault records (11.2) of the construction machinery from which the oil sample was taken, through a CRM system (11) to the database (13) (102),

[0076] - analyzing the data received from the computing device (12) and CRM system (11) in the database (13), converting it into a single format, and transferring it to a server (14) (103),

[0077] - preparing the data coming from the database (13) in the server (14) for machine learning and creating a prediction model (104),

[0078] - running an algorithm by a prediction component (14.1) within the server (14) to detect the presence of a fault and to detect the component in which the fault is present (105),

[0079] - sharing the prediction result (15) with responsible business units through the server (14) after detecting the presence and location of the fault (106),

[0080] - recording the prediction result (15) in the fault records (11.2) within the CRM system (11) after detecting the presence and location of the fault (107). The prediction component (14.1) predicts when a malfunction might occur after an oil sample is taken, using the Python programming language and preferably through a decision tree modeling library.

[0081] Based on the results from the prediction component (14.1), the server (14) categorizes them in terms of working hours, enabling maintenance predictions and the development of the service plan infrastructure.

Claims

CLAIMS1. A fault prediction system (10) for construction machinery, utilizing oil analysis values / results (12.1) for fault detection, characterized by comprising- at least one CRM system (11) containing machine / model information (11.1) and fault records (11.2),- at least one computing device (12) into which oil analysis values / results (12.1) obtained from the construction machinery are entered,- at least one database (13) that logically stores and manages machine / model information (11.1), fault records (11.2), and oil analysis values / results (12.1),- at least one server (14) for formatting and processing data coming from the database (13),- at least one prediction component (14.1) within the server (14) that enables the operation of the prediction algorithm.

2. The system (10) according to claim 1 , wherein the prediction component (14.1) processes the information in the mentioned database (13) using a decision tree model in the Python programming language.

3. A fault prediction method (100) for construction machinery, using oil analysis values for fault detection, characterized by comprising the steps of- entering oil analysis values / results (12.1) of an oil sample taken from a construction machinery into a computing device (12) and transferring them to a database (13) (101),- sending machine / model information (11.1) and fault records (11.2) of the construction machinery from which the oil sample was taken, through a CRM system (11) to the database (13) (102),- analyzing data received from the computing device (12) and CRM system (11) in the database (13), converting it into a single format, and transferring it to a server (14) (103),- preparing data coming from the database (13) in the server (14) for machine learning and creating a prediction model (104),- running the algorithm by a prediction component (14.1) within the server (14) to detect the presence of a fault and to detect the component in which the fault is present (105),- sharing the prediction result (15) with responsible business units through the server (14) after detecting the presence and location of the fault (106),- recording the prediction result (15) in the fault records (11.2) within the CRM system (11) after detecting the presence and location of the fault (107).

Images