proof read

A. A. Poddubnaya, A. V. Keller

FSUE "NAMI", Moscow, Russian Federation

E-mail: [email protected]

Abstract. The article contains general information about promising vehicle diagnostic systems. Existing diagnostic systems, including those built into modern vehicles (TS), are not able to predict the moment of failure of components and assemblies, but only state the fact of a malfunction. To diagnose the current state and forecast the residual life of the vehicle in motion mode, it is proposed to use a mathematical model based on machine learning technologies and data from standard and additional sensors, vehicle detectors. Using this approach will make it possible to forecast the occurrence of a defect before its actual occurrence.

Keywords: advanced diagnostic systems, autonomous vehicle, connected cars, unmanned vehicles, technical condition monitoring, mechanical failure detection, fault prediction, sensors, detectors, digital data processing methods

Introduction

For autonomous transport and connected vehicles, diagnostic of the vehicle’s technical condition is a basic safety standard. * The issue of determining the mechanical failure of an autonomous vehicle is extremely relevant, due to the lack of a driver who can appreciate uncharacteristic noises or external vibrations. Errors received from the vehicle’s CAN bus are not sufficiently informative in assessing the current state of the vehicle and do not predict a breakdown or a failure. For a driverless vehicle, at the stage of its design, an expanded self-diagnosis system should be laid. During operation, onboard the vehicle, data from sensors and a reliability monitoring system should be processed and further data transferred to the ITS - intelligent transport system, as well as to the servers of owners and manufacturers. (* according to researches of the European Commission.)

Main part

Almost all modern cars are modified with a variety of full-time detecting devices and sensors, fixing faults and operation errors of some nodes by electrical parameters and fixing “extreme” system states in codes. Error icons appear on the vehicle dashboard when the system diagnoses a fault. If the driver notes the incorrect operation of certain nodes, systems and you need to make sure in what, really technical condition is the transport, then a specialized diagnosis is carried out. To clarify the technical condition, the computer diagnostics of the vehicle is performed by a certified technical specialist: a scanner with software is connected to the on-board systems, through special diagnostic connectors, CAN, which reads all the codes and errors transmitted by the car about possible malfunctions on the main nodes. Error codes are currently vendor specific, are set by OEM and are available for reading and monitoring in a limited list of codes. The received codes are decrypted by specialists, again using special programs, and based on the information received, a conclusion is made about the presence of certain failures or malfunctions. On-board data consists of thousands of signals from sensors and ECUs that are transmitted through the CAN network.

They are sent repeatedly with a certain frequency and form continuous data streams that can be used both for driving a vehicle and for signaling the status of various components of the vehicle. Research on monitoring and signal analysis of standard sensors is carried out by automakers and thiere are used to date, continuous registration on board vehicles has been limited, on equipment during the testing period when developing new models or modify equipment. These systems are expensive and designed for product development.

There are not many researches in the automotive industry about resource forecasting using on-board data from standard sensors. Currently, such methods either require the participation of a person for an expert assessment or the technical condition of the car is assessed by monitoring signals and comparing them with a model of a perfect process. In the review of the development of the problem, the main approaches to the solution are formulated in the following methods: the Model Based Diagnostics (MBD) method and the Condition Based Maintenance - CBM method [1] . There are effective studies based on on-board forecasting methods:

- D’Silva carried out for a complex stationary system based on a method based on the formation of a cross-correlation matrix, including pairwise correlations between signals, where the Mahalanobis distance is used as an assessment scale to search for deviations and malfunctions [2]. The full correlation matrix is ​​used to determine vehicle status. Normal workspace is determined from experimental data. The system works on board mods with saved normal operation models, and this was demonstrated on simulated data.

- stationary signals for finding damage were used by Vachkov [3] and Kargupta et al. [4]. Their systems consist of an onboard part that continuously monitors the vehicle and loads models into the onboard analogue of OEM monitoring systems. An autonomous system includes a database in which data models, faulty and faultless systems are stored [5]. Also, for embedded on-board systems with limited resources, methods are developed in which sudden changes in the correlation matrix are signs of wear or failure [6].

Automakers until recently were not interested in developing technologies for monitoring operation, but in the present, due to the emergence of contractual relations on the principle of “full cycle service” related to the development of rental, commercial and unmanned vehicles, the topics of reliability forecasting are updated. Commercial vehicle manufacturers have not yet released any advanced forecasting solutions to the market. There are simple preventative maintenance solutions that track wear and the use of brake pads, clutches, and similar wear equipment and are predicted for the future. All of them are based on data streams that are aggregated on board and transmitted to a remote office. Mercedes and MAN, among other things, offer direct customer solutions for proactive service recommendations and remote monitoring. Volvo for commercial vehicles includes forecasting systems offered with maintenance contracts. Volkswagen, BMW [7] and GM [8] have methods for predicting future service needs based on telematics solutions and on-board data. VW and BMW offer preventative maintenance as a maintenance solution for the owner, and GM publishes recommended repairs through the OnStar portal.

Developed embedded on-board solutions have unlimited access to real-time data streams. This provides fast detection, since the detection algorithms are located close to the data source. On-board solutions usually have limited computing and storage capabilities, since the hardware must be automotive-grade, for example, resistant to water, shock and electromagnetic interference, as well as inexpensive. Typically, automotive electronics are usually two to three generations behind the consumer market.

The conceptual model developed by the authors is planned for use in terms of the implementation of remote diagnostics services of the intelligent transport system (ITS) for the connected transport. This service is assumed to be mandatory for the purpose of ensuring the operation of connected highly automated vehicles. The service will provide feedback to vehicle manufacturers on the organization of the full product life cycle.

The direction of the ongoing research coincides with the direction of work within the framework of the European Commission C-ITS. According to the act of which (Fig. 1) (Cooperative Intelligent Transport System) Delegated Act, adopted and agreed by key stakeholders from the automotive, motorcycle, agricultural, and telecommunications industries; international technical organizations are developing integrated solutions for the priority of road safety.

Figure 1. Scheme of interaction of International Technical organizations, which are making standards for the development of ITS

The 3GPP international consortium develops technical specifications and technical reports in the field of network technologies in mobile systems together with ETSI, the European Telecommunications Standardization Institute, which transfers the developed documentation on communication standards and ITS services to the ITU (International Telecommunication Union), which is a specialized institution within the United Nations (UN) and is responsible for issues related to information and communication technology. ITU transfers data to the United Sustainable Nations Development Group, the United Nations Development Group, which approves international standards at the ISO site.

In international concepts for the development of ITS, special attention is paid to the issues of diagnostics and monitoring the technical condition of connected and unmanned vehicles. The 3GPP international consortium is developing a standard for connected vehicles * 3GPP 22.885 p. 5.27 “Remote diagnosis and just-in-time repair notification” - Remote diagnosis and just in time repair notification, which provides for the installation of devices with a hardware-software complex that support interaction on a connected vehicle V2X (Vehicle-to-everything) and collect diagnostic data from sensors inside the vehicle.

Diagnostics of the technical condition of the vehicle is solved by developing a prognostic model for monitoring electrical, mechanical and hydraulic failures of the vehicle components and assemblies.

Figure 2. The relationship of databases in the developed system

The novelty of the idea and its advantages lie in the creation of a model for processing data obtained from various monitoring and diagnostic systems, using a neural network to predict the operating time to failure of a node, taking into account the optimal load mode.

Figure 2 shows the relationship of the databases of the described system, which may be applicable for various TS. Information is collected as follows: additional sensors are installed on the vehicle, data from which together with data from the CAN bus and theoretical and statistical results are received and processed on the on-board AIC by a neural network, which after initial data conversion and predictive analysis sends it to a remote ITS AIC. The mathematical models used in the agro-industrial complex rank incoming data by the integrated indicators for assessing the residual life of the diagnosed nodes. The analytical system of the ITS AIC during operation and taking into account the assessment of the impact of current loads, as well as modeling past and future operating parameters, creates a virtual dynamic operational model for a particular vehicle.

Information from sensors and detectors, based on measurements of vibration, acoustic emission and the intensity of the generated heat, determines the technical condition of the hydraulic or mechanical components of the vehicle, and is used in the construction of a virtual mathematical model on the ITS AIC. The used complex energy parameter of acoustic emission adequately estimates the change in the friction coefficient in the kinematic pair. The complex energy parameter of acoustic emission, in Fig. 3 - parameter D, is calculated based on the analysis of acoustic emission signals of a working friction pair at ultrasonic frequencies of 20-300 KHz, with ranges of operating units from 10 to 10,000 rpm. The sources of acoustic-emission signal formation in the ultrasonic frequency range are elastic deformation waves formed by the dissipation of fracture energy in the structure of materials [9]. Picture 3 shows the confirmed results obtained using an acoustic emission analyzer in experimental modeling of a friction pair.

The transition to the evaluation of the signal by acoustic emission makes it possible to evaluate the magnitude and nature of the change in energy dissipation that occurs when objects interact in the ultra-acoustic range, to obtain information on friction in the assembly and to diagnose this process with an assessment of the state of the friction pair, landing quality, nature and lubrication conditions and a number of other parameters in contact. So, when changing the level of the acoustic emission signal in the friction unit, it is possible to evaluate the lubricant quality of the controlled unit. Constant monitoring of the quality of the lubricant will reduce the wear rate and the development of defects.

To assess energy losses by the acoustic emission method, a diagnostic tool is used, which fixes the integral indicator of the complex energy parameter of acoustic emission. According to the results of previous studies, RF patent No. 2427815 - G01M13 / 02 test transmission mechanisms "Method for the diagnosis of mechanical transmissions" [10].

Figure 3. The ratio of the complex energy parameter D to the force at the contact spot of the simulated kinematic pair, N

Calculation methods have objectively established the proportionality of the integral indicator of the acoustic emission sensor signal and the friction coefficient, both for gears of gears and bearings. It is this area that has expanded the capabilities of existing non-destructive testing methods, will allow solving practical problems of monitoring and forecasting the state of technology, rationally distributing forces and means during repair and inspection of the technical condition of a vehicle. The resource forecasting in the traditional approach is based on the design and construction parameters of the elements of the nodes, and failure statistics and does not give a real result on operational reliability. It is proposed to change the approach to data processing methods using machine learning. In this case, node element failures are considered as some abstract random events of the multifactor process, and the diverse physical conditions of products are reduced to two states: serviceability and malfunction. Prediction problems must be considered with errors in the initial and boundary conditions even when the non-stationary process can be considered as a strictly deterministic process, that is, its outcome is completely determined by the algorithm, the values ​​of the input variables and the initial state of the system. Based on mathematical models of vehicle components and the dependences of changes in the data of sensors, detectors when compared with operating conditions, as well as data on the CAN bus, it is proposed to create an analytical complex applicable to a wide range of equipment and vehicles.

The resource forecasting in the traditional approach is based on the design and construction parameters of the elements of the nodes, and failure statistics and does not give a real result on operational reliability. It is proposed to change the approach to probabilistic data processing methods. The calculation is built on the probability of a node element failure depending on the current state of the interface and loads. In this case, node element failures are considered as some abstract random events of the multifactor process, and the diverse physical conditions of products are reduced to two states: serviceability and malfunction. Prediction problems must be considered with errors in the initial and boundary conditions even when the non-stationary process can be considered as a strictly deterministic process, that is, its outcome is completely determined by the algorithm, the values ​​of the input variables and the initial state of the system. Based on mathematical models of vehicle components and the dependencies of changes in sensor data, when compared with operating conditions, as well as CAN network data, it will be possible to create an analytical complex applicable to a wide range of vehicles and vehicles.

The complexity of the implementation of the diagnostic complex lies not only in the need to take into account the operating conditions of a particular vehicle, but also to carry out maintenance in accordance with these conditions. A modern service system should interconnect the data of continuous monitoring of the technical condition of individual vehicles; planning of their operation by the trucking company and the willingness of the service department to fulfill the required list of maintenance and repair of vehicles.

A common problem that the operating organization faces when using service schedules and planned site replacements is the performance of an inappropriate amount of asset maintenance. Since calendar-based maintenance does not take into account the asset’s performance, the frequency of maintenance work can often be either too high or too low. These problems can be prevented by optimizing and improving preventative maintenance programs.

The system of preventive diagnostics based on artificial intelligence will make it possible in the future to abandon the system of scheduled preventive repairs and switch to maintenance by actual condition [11], which experts estimate will reduce maintenance costs by 75%, the number of services by more than 50% reduction in the number of failures by 70% for the first year of operation.

Conclusions

Advantages of maintenance using the developed system:

· The system can be applicable in various industries, the system is adaptable;

· Registration of signs of fracture appearances and analysis of their variations is performed. Assessment of the probability of failure of both the system as a whole and its individual nodes;

· Failures of nodes are transferred from the sudden category to the predicted category, due to early detection and notification of personnel about a developing malfunction;

· Spare parts logistics is being optimized - timely ordering and delivery.

The expected economic effect from the introduction of the developed system of preventive diagnostics is possible due to reduced maintenance costs, reduced downtime due to malfunctions, and repair costs will be reduced. The developed system solves the problems of insufficient competence of employees in assessing the technical condition, and will also help maintain the working condition of aging equipment in a limited budget.

Realized analogues in the aspect of objective resource forecasting do not exist. The advantage is the prospect of integration into the ITS (Intelligent Transport System), and the creation of a predictive service for any type of vehicle, and most importantly this is the only way to diagnose unmanned vehicles in the field.

References

[1] Fault-Diagnosis Systems - An Introduction from Fault Detection to Fault Tolerance, Authors: Isermann, Rolf - ISBN 978-3-540-30368-8

[2] S. H. D’Silva. Diagnostics based on the statistical correlation of sensors. Technicalpaper 2008-01-0129, Society of Automotive Engineers (SAE), 2008

[3] G. Vachkov, “Intelligent data analysis for performance evaluation and fault diagnosis in complex systems,” in IEEE International Conference on Fuzzy Systems, - July 2006, pp. 6322-6329

[4] H. Kargupta et al., “VEDAS: A mobile and distributed data stream mining system for real-time vehicle monitoring,” in Int. SIAM Data Mining Conference, 2003.

[5] Mohammad MesgarpourDario Landa-SilvaIan Dickinson - Overview of Telematics-Based Prognostics and Health Management Systems for Commercial Vehicles - 13th International Conference on Transport Systems Telematics, TST 2013, pp 123-130

[6] Hillol Kargupta, Michael Gilligan, Vasundhara Puttagunta, Kakali Sarkar, Martin Klein, Nick Lenzi, and Derek Johnson. MineFleet®: The Vehicle Data Stream Mining System for Ubiquitous Environments, volume 6202 of Lecture Notes in Computer Science, pages 235-254. Springer, 2010.

[7] Von Harald Weiss. Ingenieur.de predictive maintenance: Vorhersagemodelle krempeln die wartung um. https://www.ingenieur.de/technik/forschung/predictive-maintenance-vorhersagemodelle-krempeln-wartung-um/, 2019-08-23.

[8] OnStar. On Star on-star services. https://www.onstar.com, 2014, 2018-08-23.

[9] - Assessment of the lubricity of gear oils of mining machines -A. A. Poddubnaya, A. S. Fokin, S. L. Ivanov, E. A. Kremcheev, V. S. Potapenko - Notes of the Mining Institute, vol. 178 - http://pmi.spmi.ru/index.php/ pmi / article / view / 2176

[10] - RF patent No. 2427815 - G01M13 / 02 test transmission mechanisms "Method for the diagnosis of mechanical transmissions." http://www.freepatent.ru/patents/2427815

[11] - Management of equipment maintenance and repair: automation capabilities - I.N. Antonenko NPP SpetsTek 2019-05-20, http://trim.ru/sites/default/files/files/pdf/maintenance_food_industry.pdf

7