The modern transformation of the logistics outlook increasingly changes the world and sets new tasks in the development of information technologies for production management and infrastructure operation. This calls for the development of digital and intellectual technologies for the day-to-day management of capital- and resource-intensive enterprises: railway and automobile transportation, automobile construction and locomotive construction, depot, and maintenance shops.
A huge amount of data, generated by industrial units, has already been accumulated. Thus, it is possible to apply highly effective predictive tools to rapidly and qualitatively evaluate the condition of operating units, detect the origin and development of abnormalities and identify trends that lead to unscheduled disruptions and failures, by registering the failure probability and remaining service life in real time.
The idea of accurately predicting the technical and qualitative condition originated from the need of resource- and capital-intensive companies to reduce the fast-growing expenses of owning equipment, advertising services and increasing competitiveness.
A big boost has been given to the development of intellectual railway systems by modern tools, such as hybrid models, explainable AI, Industrial Internet of Things, big data, cloud computing and machine learning. These technologies enable optimization of transportation resources, thereby enhancing transportation efficiency.
1,2
Range of Problems – Locomotives, Wheeled and Air Transport
In the elevated uncertainty conditions, a travel supplier is forced to keep non-production reserves of haulage resources against the possibility of unpredictable growth of the transportation load.
For example, for a Russian railway company, excessive off-road time for scheduled repairs, as well as running repairs and waiting for free repair facilities, led to the necessity to take out of service over 1,000 locomotives a day.3 It has been established empirically that a necessity exists in a locomotive that is ready for service 90 percent of its lifetime during the whole lifecycle.
Figure 1: Statistical locomotive time budget, 2016
Electric Trains
When failures of commuter electric trains occur and upset train time schedules, they dramatically reduce the service level. A breakdown of a single train upsets time schedules of other trains. A seamless movement is dependent on the technical condition of each train.
The basic risk is an episodic falling of railroad cars out of traction. This occurs when a railroad car is disengaged from the traction of the train for some reason (e.g., breakdown, protection systems), but traction is continued by the remaining power railroad cars. This undermines the operation of traction units of all power railroad cars. Predictive analytics make it possible to provide the respective information in advance, before a power railroad car stops to be engaged in the traction.
For example, under normal operating conditions, railway electric motors should equally contribute to the total traction of the electric train. Consequently, considerable variations of energy consumption parameters of individual railroad cars will indicate a malfunction in systems of that railroad car.
In Saudi Arabia, the situation of a train failure amid a desert is of great risk. In such cases, labor inputs and the duration of repair work increase. Subhuman conditions for the passengers are created, who are forced to stay in a frosty railway car for a long time. Prediction and prevention of such breakdowns are critical in such conditions.
Modeling and Condition – Based Maintenance
Mathematical modeling of complex production facilities is fragmentary in nature because of:
- A lack of high quality, telemetric data from equipment;
- A lack of calculation methods and idealized mathematical models for complex systems;
- Insufficient modeling of individual units;
- A lack of academic privileges in three fields: MRO, mathematics and IT.
Despite more and more scientific articles published over the last few months where scientists are able to combine mathematical modeling methods with machine learning, most of these articles are oriented toward situations that have either a sufficient theoretical development of parts of the modeled system or a sufficient volume of high quality data. The experience shows that such “hothouse” conditions are extremely rare in the world.
Solution to the Problem
An early detection of emerging defects and failures during operation makes it possible to diagnose problems before they turn into accidents. For example, if a deviation is recorded even before a parameter comes to the pre-alarm level, then a possibility exists of localizing the defect quickly, planning the logistics of spare parts and other resources, and carrying out a scheduled repair.
Predictive analytical solutions collect information from sensors in real time and use it together with the historical data, visual inspection, manual measurement data, video records, equipment functioning data and other useful data.
To solve the aforementioned complexities, one needs to use up-to-date methods for building prediction systems, knowledge bases and expert systems. As a result, a complex of hybrid systems has appeared that holistically describe the behavior of complex systems, both technical and technological.
Smart Locomotive – Implementation Example
Since 2017, a locomotive technology company in Russia has been working on a “smart locomotive” project with Russia’s state-owned railway company, which manages infrastructure and operates freight and passenger train services. The rail company maintains and repairs about 70 percent of the entire Russian locomotives and has 92 depots and 10 repair factories.
A system for intellectual diagnostics and technical condition prediction of locomotive equipment has been developed. The failure search module has been implemented on 4,000 locomotive sections. Now, it is possible to identify over 50 types of equipment disturbances and monitor over 20 types of various equipment: traction generator, traction electric motor, fuel and oil pumps, water cooler, turbine compressor, dynamic brake, etc. The system is integrated into the company’s enterprise resource planning (ERP) system. A workshop order is automatically issued to carry out repair work based on detected faulty operation data. This makes it possible to calculate required repair resources and prematurely update the locomotive-to-depot schedule for both planned and unplanned maintenance.
To date, the results show:
- Enhanced reliability and safety of locomotives running on lines: Failures on the line are reduced by 32 percent.
- Improved operational efficiency of the process: Locomotive diagnostic time is reduced from four hours to 10 minutes because the diagnostician gets all the failure incident data before the locomotive enters the depot.
- Increased economic efficiency of the process: The amount of fines imposed on the locomotive technology company by the railway company for failing to provide the technical availability coefficient dropped by more than 20 percent.
The smart locomotive system identifies impending equipment failures weeks or months before they happen. This valuable information gives the railway the opportunity to transform maintenance into a condition-based process.
The solution uses telemetry data from an onboard data transmission system (referred to as BDTS), a diagnostics system and an ERP system. In addition, it uses information about external factors, such as weather, violation of operating modes, etc., to make predictions more precise.
The diagnostics system detects anomalies, as well as operators’ incorrect actions, by means of a mathematical model and then sends them to the ERP system. An anomaly is a situation where any important equipment parameter value differs from the normal one predicted for current operating mode. All the anomalies detected in a server are verified through workstations.
The system also trains the models using feedback data from the ERP system, based on information about performed work and components replaced in the process of repair operations.
Comparing the rate of change in absolute and relative parameter values makes it possible to determine the equipment degradation rate. Therefore, the diagnostician receives all necessary information about the condition of the railway rolling stock system at least 100 hours before the locomotive enters the depot.
Figure 2: Algorithm of a "Smart Locomotive"
Conclusion
Predictive solutions help suppliers build more productive mutual relations with the users, thanks to receiving efficient information for making decisions. Such products enable suppliers to predict the lifecycle of their equipment and help the users understand the condition of their equipment and control performance of guarantee obligations. The understanding of the full picture helps the parties conclude correct contracts and find the most efficient interactive ways. For example, a major truck manufacturer in Russia, having placed stake on the predictive analytics development and introduction thrust, totally rebuilt its business models. The company has begun moving from the supply of transportation facilities to the sale of operational kilometers or cargo tonnage haulage up to the conclusion of lifecycle contracts.
The practical use of such hybrid models enables a considerable reduction in fleet maintenance and minimizes the number of unscheduled repairs, thereby improving the operational reliability of the MRO system.
Such a system combines in itself a functionality for integration of data, analytics and supporting tools for decision-making. It all has been developed as a unified, adjusted software product that can be easily used in most of IT structures.
References
1. Rail Safety and Standards Board Limited. “Internet Access on Trains for Customer and Operational Railway Purposes.”
Rail Industry Standard RIS-0700-CCS: Issue: One, June 2016.
2. Obukhov, A.D. “Digital technologies in management of operation activities in the railroad transport.”
Automatics, communication, informatics: 2017, No. 9, p 4-8.
3. Valinsky, O.S. “Improving the efficacy of locomotive complex management.”
Locomotive: 2017, No. 1, p 3-7.
