Este site utiliza cookies. Ao continuar navegando no site, você concorda com esse uso. Leia nossa política de privacidade>

Se precisar de ajuda, clique aqui:

Xi'an Locomotive Depot Safeguards Railway with AI & 5G

By Guo Qianli, Chief Architect of Railway Industry, Global Transportation Business Unit of Huawei's Enterprise Business Group; Chen Danfeng, Senior Architect, Global Transportation Business Unit of Huawei's Enterprise Business Group

Tasked with ensuring the safety of train operations, locomotive experts are invaluable in railway transportation. Huawei's Smart Locomotive Solution, which is powered by 5G and AI, can free these key workers from exhaustive, cumbersome tasks and significantly improve safety assurance capabilities and efficiency — as railway transportation sets a standard for upgrading industries from person-assisted to tech-assisted operations.

China's railways supported 3.66 billion passenger trips in 2019. The safety and comfort of the passengers on each of those train rides were ensured by locomotive experts — including several specialists.

Several of those specialists are deployed at the locomotive depot at China Railway Xian Group (CR-Xi'an), which is responsible for all passenger trains departing from or passing through Xi'an — the capital city of northwest China's Shaanxi Province, with a population of about 12 million.

Those specialists include Li Xiangqian, the Chief Engineer of China Railway, who receives a special allowance from the State Council of China; Zhou Miaosheng, who holds a National May 1st Labor Medal and has been officially recognized as a role model of CR-Xi'an and as an 'admirable railway employee during the Spring Festival travel rush'— a time when hundreds of millions of train journeys are made nationwide.

Other experts at the depot include image analysts, such as Sang Minghui, who is able to gain insight from typically under-used data and makes decisions based on video footage.

These experts are busy year-round, but their workloads increase sharply during the Spring Festival travel rush. The amount of work to be done is so vast that these experts are typically only able to spend Chinese New Year's Eve with their families every four or five years.

As well as being responsible for all passenger trains departing from or passing through Xi'an, CR-Xi'an's Locomotive Depot also monitors the performance of train drivers (including drivers of high-speed railways). Locomotives and drivers are critical to ensure safe train operations, so locomotive experts start working as soon as a locomotive enters the depot, which is when image analysts monitor drivers' operations and specialists check the health status of locomotives.

Manual Operations Limit Image Analysts' Capabilities

Because they are directly responsible for the property and safety of thousands of passengers, train drivers face a lot of pressure. Tasked with controlling safety risks, Xi'an Locomotive Depot has assigned 40 dedicated image analysts to perform correlation analysis of video footage and train control and monitoring (LKJ) system data while trains are en route to their destinations, as well as do spot checks and record driving violations, and then implement appropriate sanctions or penalties.

Typically, each locomotive is equipped with seven cameras to record videos of the cab, mechanical room, and rail conditions when the vehicle is in motion, generating about 30 GB to 50 GB of data each time it performs an operation. A locomotive's onboard data is manually copied using USB flash drives, and the data transfer takes over 40 minutes, which is inefficient and labor-intensive.

Complicating things, manual data copy can cause various safety risks, for example, the onboard computer may be infected with the USB flash drive viruses; difficult management and control of multiple intermediate links pose data tampering and damage risks; and the USB ports of onboard devices may be destroyed.

The locomotive domain suffers from a lack of informatization and intelligence, which has severely limited its service development.

Manual video analysis can't cover massive amounts of video data and problems are easily overlooked when analysts become fatigued and their eyes are tired. Another problem is that experienced train drivers' expertise is wasted as they need to act as image analysts.

The Urgent Need for Tech-Assisted Methods

A total of 70 specialists from Xi'an Locomotive Depot's dedicated inspection and repair team continuously check the health status of hundreds of locomotives every day. They climb on top of trains, go down to trenches, and walk around in narrow mechanical rooms. All kinds of specialists — locomotive and electrical fitters, and braking, welding, or pantograph- Overhead Catenary System (OCS) system detection engineers — need to comprehensively check each locomotive in the shed within a limited time period (20 minutes during the Spring Festival travel rush), quickly locate faults, and act fast — ensuring all locomotives can drive out of the shed on time and run safely.

Integral to locomotive efficiency and safety, running gear is roughly equivalent to a car chassis, though it's more complex. It's the lower part that guides the train to run along the tracks and transmits all the weight to the rails. It comprises the wheelset, journal box lubricators, side frames, bolsters, and spring shock absorbers.

A single set of running gear usually includes 2306 screw bolts. Using the conventional repair and examination mode, railway staff need to check whether each of the screw bolts and other key components of the running gear are working properly — either by using only their naked eyes, or by tapping on the components with a hammer and listening to the sound. This relies on the railway staff having a lot of experience of manual checking, and even that doesn't guarantee that all components will be checked effectively. In addition, workers need to perform multiple complex operations for each locomotive after it moves into the depot. Although various sensors are installed at key components, the sensor data isn't fully used, so frequent manual maintenance is required, and this wastes both manpower and material resources.

When the key components have major health issues, specialists need to take targeted measures based on the locomotive's repair process, historical problems, and symptoms, and send the locomotive to the shed or factory if necessary. More systematic and comprehensive diagnosis and specialized treatment can take weeks, which is severely detrimental to the operational efficiency of railways. Some of the more complex issues rely on specialists' experience. The transition from person-assisted to tech-assisted repair, and consistent optimization of health management for locomotives and key components, as well as repair classes and systems need to be implemented as soon as possible.

Smart Locomotive Solution: Powered by 5G and AI

After years of involvement in the industry, Huawei has gained in-depth insight into railway transportation. Applying its leading communications and digital platform capabilities, Huawei and industry partners launched the innovative Smart Locomotive Solution, which uses 5G and AI. This solution reduces locomotive experts' heavy workloads, enables more intelligent operations, and better ensures railway transportation safety — helping CR-Xi'an embark on a smart journey.

This solution uses Huawei's world-leading 5G technologies to replace manual operations. It can lead to a ten-fold increase in data transfer efficiency and automatically obtain real, complete driver surveillance footage.

The solution works on the emerging mmWave band and uses beamforming (phased-array antennas, supporting fast attenuation and low interference beyond the coverage area) and intelligent tracking (ensuring the alignment between the Railway Base Station (RBS) and Train Access Unit (TAU) at all times, with stable bandwidth) technologies, to achieve high-speed train-to-ground data transfer. When a locomotive slowly enters a depot, a high-speed channel of 1 Gbit/s or above is automatically established between the TAU and RBS within 300 to 500 meters in the depot's throat section, supporting a maximum transfer rate of over 1.5 Gbit/s. About 30 GB of onboard data is generated during each locomotive operation, which can be automatically dumped within three minutes.

The entire data transfer process ensures data integrity, and it's secure and reliable without manual intervention, ten times faster than manual data copy. The smart locomotive solution is the railway industry's first intelligent application of 5G. It will facilitate more efficient manual data copy, eliminate safety risks, improve data reliability, and tackle many other issues, providing complete video data for image analysts.

A comparison of conventional and smart locomotives

The smart locomotive solution has raised the detection rate of image analysts. It uses 5G to transfer all driver-related video data to the intelligent analysis center through high-speed cache devices, and automatically identifies violations using AI.

Huawei's AI platform — HUAWEI CLOUD ModelArts — provides industry-leading computing power, enabling partners to shorten the training duration of a single model from one week to only one to two days and supporting fast application. It is the fastest training platform certified by Stanford DAWNBench — a benchmark suite for end-to-end deep learning training and inference — setting a new record of 4 mins and 8s (it takes 18 minutes to perform the same type of model training on the AWS).

The solution also supports customized development of video-based behavior analysis algorithms based on actual scenarios. A total of 11 driving violations can be automatically, intelligently identified, such as not making gestures, using electronic devices, not operating the steering wheel or workspace equipment properly, and showing signs of fatigue. The intelligent driver evaluation mode helps regulate driving behavior and ensure safety, eliminating the issues caused by the inefficiency and insufficiency of manual spot checks. In the past, one employee could only check four locomotives every day; with the intelligent analysis technology, an employee can process analysis results of over 40 locomotives daily, increasing the efficiency by ten times.

The smart locomotive solution uses big data diagnosis to predict the health status of key components. Using the Huawei big data platform's integration, storage, query, analysis capabilities, as well as support for data warehouses and Business Intelligence (BI), and data connectivity and sharing capabilities of the ROMA platform, the solution streamlines multiple core service systems to build an End to End (E2E) locomotive information sharing platform.

The solution builds fault prediction and health management models for locomotive running gear based on the data of onboard sensors or concerning ground operational safety, servicing, and repair, as well as extensive expert experience and fault examples accumulated by Huawei partners. Such models implement heath evaluation and service life prediction functions, and they provide a scientific basis for the reform of repair processes and systems during the transition from planned to predictive maintenance.

The big data platform supports analysis of massive amounts of historical status data of running gear sensors and health evaluation of key components. By drawing endurance curves and generating timely alarms for unhealthy components, the platform minimizes manual inspection and repair times and prevents minor issues from escalating into more difficult problems.

Big data can also be applied to detect flaws in railway axles. The health status of axles is crucial for safe train operations. Inefficient detection of axle damage causes larger and larger cracks, which can lead to disastrous consequences such as train derailment. The damaged key components, such as drawbars and couplers must be replaced in time for safety reasons. In line with regulations, any wheelset with over 2-mm holes or 0.7-mm peelings should immediately be replaced. However, damage to axles is usually very minor, and it usually takes three or four rounds of manual inspection before the damage is identified and confirmed. With big data analysis, the solution obtains the status data of axle sensors — such as temperature and vibration — to evaluate whether they are healthy, and predict fault risks and generate alarms accordingly. As well as increasing the repair efficiency, it also reduces safety hazards.

Smart Locomotive: From Person-Assisted to Tech-Assisted

The application of Huawei's 5G- and AI-based Smart Locomotive Solution reduces the safety issues that occur while locomotives are running by more than 10% and leads to an over ten-fold increase in the data collection and analysis efficiency, saving millions of CNY (CNY1 million is about US$153,000) in operational costs per locomotive depot per year and greatly reducing safety risks.

The solution allows locomotive experts to spend time with their families during national holidays, while safeguarding the travel of the potential 1.4 billion train passengers who live in China.

ICT Insights on the go

Scan for mobile reading

ict 30 400 514

Issue 30 —Smart Transportation Special

Bridging the IT and OT Digital Divide with a New Digital Transportation Paradigm

Current catalogue

Comment

Features

Special Reports

Focus

Success Stories