Predictive Maintenance for Escalators in DPP, Prague’s Public Transport Company

Prague subway transports a million passengers every day and the escalators are essential. So the mechanical components of the escalators are constantly used and worn. The operator needs to know when maintenance and replacement of the damaged component are required to avoid their failure. Neuron soundware has a predictive maintenance solution for such a problem.

The goal of the project is to minimize the number of failures and escalators downtimes. Thanks to the system information handed to the service team the service time and its costs are lowered. Automated digitized solution for a long time machine maintenance expands the lifespan of the machine.)

nShield, the customer software, displays the escalator behavior analyzed using audio and AI in real-time. The data is collected by IoT HW solution with acoustic sensors.

Based on mathematical models the system can identify the anomaly and the deviation of the component from the nominal state. When we apply the knowledge of the mechanical component the system can then identify the particular failure and notify an operator with a recommendation to particular maintenance or displacement.

Blog
Maintenance operator gets an app with a clear overview of the individual components. Including diagnostician insights and recommendations for diagnostics on particular components. The access to the information is remote and online.

What DPP representatives say about the solution

“Like all machinery in the transport system, escalators can break down, which can lead to delays in a critical part of the transport infrastructure. We want to embrace digitalization in monitoring our equipment, which is why we chose a solution where sensors collect acoustic data and process it using AI. This allows us to monitor the equipment remotely and send a specialist out if our Neuron Soundware equipment warns us of a change in the condition of the escalator parts,” comments Ing. Petr Vondráček, Head of Transportation System Service, Traffic Route Metro.

How does a founder of NSW perceive such a unique project with DPP

Difficulty accessing the equipment, needing to make repairs outside ordinary operations, and long waiting times for replacement parts all make escalator maintenance more complicated. Neuron Soundware is fitting 21 escalators with 189 sensors, adding the Prague transport company to its list of customers using automatic machine monitoring in the automotive, mining, and energy industries.

“Certain escalator components are found in places difficult for a human to access, often with no phone signal or internet connectivity. For cases such as this we developed a version of our equipment that allows us to process the signal completely in the end unit, at the location where the sound is recorded. We make use of a reliable, ultimately less costly data processing solution on-site in a microcomputer with no need to send terabytes of data from each machine to the cloud,” comments Ing. Pavel Konečný, CEO at Neuron Soundware.

What the project looks like from installation to commissioning

HW IoT solution installation phase

Escalators are equipped with a sensor group that records audio samples from monitored machine components to the nBox central unit. Data is sent via LTE to be processed and saved in the customer’s portal.

Acoustic data collection phase & monitoring

Sound samples from escalator operation are stored in short intervals between 24/7 to Cloud.

Artificial intelligence monitors escalator sound outcomes and evaluates this data against the escalator nominal behavior. The operators have a complete overview of the machine condition. 

Alert phase: neural network machine learning for collected data evaluation

Recorded data are evaluated by trained neural networks and visualized on a web portal. Here DPP operators can access the data. At the same time, they receive email and SMS information alerts of any changes in sound on specific components. So the operators can prioritize and plan their maintenance activities. 

The phase of operation, evaluation and action recommendations

The alert is evaluated against the machine library and diagnostician expertise. The operators receive recommendations and navigation on the machine and its component action.

The phase of taking the service to the next level

After recorded anomaly evaluation we upgrade the algorithms so the service gets significantly improved. Next time the failure occurs the artificial intelligence detects the emerging anomaly sooner and with greater precision. The ROI of the whole service is increased and new machine behavior knowledge is updated.