Open Data Platform for Lift Optimised Service: An Opportunity for SMEs in the Lift Industry

by David Abadía, Ricardo Salillas, Alfredo Gómez, Ricard Bou, Pedro Fernández, Lorenzo Beltrán

This paper was presented at the 2022 International Elevator & Escalator Symposium in Barcelona, Spain.

 Abstract

A consortium made up of Ascensores Beltrán, MP Ascensores, Nayar and ITAINNOVA is working on the development of an open platform for lift optimised services. The project aims to develop an advanced data analysis platform for optimising elevator maintenance and service based on knowledge extracted from the combined data of connected elevators.

The platform will be based on a digital infrastructure and software modules by integrating machine learning algorithms and advanced analytics that will work on the lift data entered into the platform. The outcomes of these algorithms will be aimed at reducing service costs based on deeper knowledge about the status of the lifts obtained by means of the identification of evolution or wear trends, the detection of operating anomalies, the prediction of behavior or anticipation of alarms.

The main characteristic and differentiating element of the proposed approach is that the platform must be open to the user (elevator maintenance and service companies) and to the developers of analysis algorithms. In order to provide data compatibility, data standardisation will be applied. This paper presents the considered objectives, technical approach and architectures initially designed for the platform.

The project presented in this paper is the result of the combined effort of several companies belonging to AECAE, the Spanish Association of Lift Components Manufacturers: MP Ascensores as a global company in the area of lift manufacturing and installing, Ascensores Beltrán as a small and medium-sized enterprise (SME) manufacturer and installer of bespoke lift solutions, Nayar as technology manufacturer of IoT solutions for lifts, ITAINNOVA as an R&D center and AECAE itself as coordinator and sponsor.

1. Proposed Approach

The project aims to develop an advanced data analysis platform to optimise lift maintenance and service tasks based on the knowledge extracted from the combined data of connected lifts in terms of their operating status.

• Advanced data analysis platform: Digital infrastructure that serves as the basis for the operation of certain software modules with which it is compatible. These software modules will consist of implementations of machine learning algorithms and advanced analytics that will work on the data entered into the platform.
• Optimisation of maintenance and service tasks: Reduction of service costs based on greater knowledge, also obtained in real time, about the condition of the lifts.
• Knowledge extracted regarding the state of operation: Identification of trends of evolution or wear, detection of operating anomalies, prediction of behavior, anticipation of alarms, etc.

The main characteristic and differentiating element of the proposed approach is that the platform must be open in the sense that it can be used by multiple actors with different characteristics and requirements.

Thus, because it is considered open from the user side, elevator maintenance and service companies will be able to enter the data collected from their connected elevators in an anonymised manner on the platform and obtain relevant information from the same to optimize their maintenance services.

In the same way, an open platform is also considered from the side of the developers of analysis algorithms who will be able to implement them in the platform offering new analysis services that give rise to new features.

This approach introduces certain complexities from the point of view of compatibility with multiple types of data and use cases that may arise that will be resolved through the standardisation of data formats and protocols.

Thus, a basic architecture is proposed based on connectors (public application programming interfaces (APIs)) between the data platform and the user’s data platforms or repositories in such a way as to guarantee control over them by their owner.

On the other hand, several complexities are also introduced from the point of view of platform management given the multiplicity of agents involved, both on the user side — elevator maintenance companies — and on the side of algorithm developers — technology companies. These complexities will be resolved through a management and business model based on PaaS (Platform as a Service) models.

Based on this scenario, a general approach is proposed in Figure 1.

The platform is made up of the blue boxes, with open APIs, which store normalised data and results and can generate graphs, reports, indicators and events.

The yellow boxes represent the user data that reaches the platform through APIs and cover their use case using graphs, reports, indicators and events delivered by the platform also through APIs, or by defining their own.

The green box focuses on the analysis algorithms that are also connected to the platform to generate results and that may be developed by different technological agents under the platform’s management model and standards.

With this architecture, the user will add their data and obtain specific service reports for each elevator, by company and at an aggregate level. This is information intended, on the one hand, for the route technician and, on the other hand, for the company’s management.

2. Strategic Objectives

With the proposed approach, the following goals are pursued at the association and sector level to:

offer the possibility to small elevator maintenance and service companies (more than 95% of the total in Spain) to access advanced analysis tools for the optimisation of the service that they could not develop by themselves nor would it make sense given the size limited maintenance portfolio of these companies.

generate a platform model from which more value is progressively obtained as more users add data to it, while guaranteeing their security and confidentiality. The target scope will, therefore, be national and international in later stages.

attract algorithm developers in such a way that users can choose between different alternatives and services according to their interests, generating, at the same time, “competition” between the analysis algorithms so that improvement cycles are generated.

3. Technological and Business Objectives

At a technological level, the creation and implementation of the following fundamental elements will be required:

• IoT systems to provide connectivity to lifts, including communication elements and customer data repositories (customer platform)
• API-Interfaces for connecting the client platform with the analysis platform
• Cloud infrastructure
• Machine learning and data mining algorithms for processing and calculation
• Developments for information security and integrity

At the business and management level, derived objectives are the creation of:

• A pay-per-use model that guarantees transparency and reliability to users and algorithm developers
• A management model that guarantees and facilitates the incorporation of new users and developers to the platform

4. State of the Art

The industry is experiencing an increase and accumulation of data never seen before. These data comprise a variety of formats, with different semantics, typology and quality, and often come in real time, such as, for example, sensor data from equipment (lifts), machine parameters, production lines and environmental data.

This phenomenon, called Industry 4.0,^[1] increases the availability of large amounts of data coming from the Internet of Things (IoT),^[2][3] leading us to a new world called Big Data,^[4] which allows the information they contain to be analysed and exploited to obtain added value. New developments in certain domains, such as mathematics and computer science, offer great potential to transform the industrial and building environment by understanding and exploiting this amount of process data.

One of the most exciting developments is in the area of machine learning,^[5] a subset of the field of AI,^[6] which consists of seeing how to make machines learn autonomously, based on historical data. The great applicability and the low barriers for the development of AI allow us to undertake innovations that were previously thought unattainable.

Lifts transport passengers constantly and, at the same time with each of the trips made, a large amount of information is generated, which can be analysed through AI for the improvement and exploitation of the service. Intelligent data analysis (deep analytics, applied to large amounts of data, Big Data) through AI allows the generation of models that learn how the systems of an elevator installation work and what patterns identify operating behavior and breakdowns, considering the correlations between them and their evolution over time.

Among AI technologies, deep learning^[7][8][9] stands out, positioning itself as one of the most innovative and powerful techniques for getting computers to “learn” in a similar way to humans, trying to simulate the complex behavior of the human brain and its ability to recognize patterns through the sensory stimuli it perceives.

Through deep learning, as well as other classic AI techniques, for example, analysis of time series of the evolution of variables, the extraction of behavior patterns with a high level of abstraction, trend analysis and behavior prediction and identification of relevant variables for the characterisation of processes can be performed, among many other applications.

The availability of data from sensors describing lift operation parameters in combination with AI offers great potential to improve processes in the field of lift service, making possible the analysis of behavioral trends in installations, in short, extract knowledge from heterogeneous data collected from the lift combined portfolio.

Some different application possibilities from the extracted knowledge are:

• Monitoring and detection of anomalies and faults: The acquisition of data in real time and its deep understanding through machine learning techniques allow the characterisation of the behavior of a lift installation. This knowledge makes it possible to exploit the data of an installation, as well as detect anomalous patterns in real time. For example: divergence in terms of kinematic and dynamic parameters, maneuvers carried out and parameters of the installation components allow correlations to be established with system faults that would not otherwise be detectable, thus allowing the early identification of anomalies in the system and anticipating avoiding economic consequences and service stoppage.
• Predictive maintenance: AI allows effective modelling of the behavior of the different components involved in a lift installation. In this way, machine learning models learn how the components operate, how they deteriorate and what failure patterns can occur, allowing data to be analysed in real time to predict when a repair action must be carried out preventively to avoid damage. An operation failure takes into account normal operation patterns and breakdowns, as well as the use made of an installation. This anticipation enables a preventive resolution of the problem without affecting the quality of the service.
• Optimisation of the operating parameters: It is based on the understanding of the behavior of a lift installation itself and the variables of the environment, which may be present such as users, climatology, to allow an optimal configuration of the parameters of operation taking into account the traffic patterns and behavior of the lifts. For example, by identifying user behavior patterns based on use, trajectory and schedule, the most common patterns possessed by residents of a building can be determined, allowing optimisation of use and improvement of its quality of service.

With these objectives, the large multinationals in the sector have developed advanced solutions in this field that are already being exploited commercially and that are unbalancing the competitive landscape even more, taking into account the difficulties that an SME must face in order to develop an advanced maintenance system, optimised, both for technical capacity and for data availability when handling much smaller maintenance portfolios.

For an SME, even a medium-large one, it is very difficult to develop solutions of this level due to both limitations in technical capabilities and the volume of data required to make it technically feasible.

This project aims to address these limitations based on cooperation between the promoting agents and fundamentally based on the proposed open platform approach.

5. Technical Description

The approach and technical architecture planned for the platform is described below based on the general scheme presented in Figure 2.

Architecture Components and Workflows

Lift Infrastructure (Green Box)

It includes all the components of the lift infrastructure (car, control panel, available sensors, button panel, etc.), as well as a node that allows the information coming from these data sources to be collected, processed and prepared for sending, and a communications module that sends the information to the infrastructure of each elevator operator.

The identified components are:

• Lifts (elevator infrastructure): control panel, car, available sensors, button panel, etc. (understood as data sources). It is important to note that the events of the control panel that allow reconstructing the journeys made must be collected (for example: opening /closing of doors, passage through the floor, movement, button press … with its corresponding temporary stamp).
• Elevator platform (processing node): It is responsible for accessing all the information on the lift installation, collecting it, preparing it for shipment, identifying the lift, the installation.
• Elevator platform (communications module): It is responsible for sending the information collected through secure communications to the elevator operator’s infrastructure, sending the information collected:
  • Not necessarily in real time, it can be done at a frequency that allows communications to be optimised (for example, once a certain amount of information is collected, either by time or in terms of size).
  • It must include the information of the events that allow reconstructing each of the travels and variables of interest (for example, door opening time, door closing time) with the time stamp of each of them, as well as those sensors relevant to analyse (for example drive power).
  • Secure communications must be provided according to the infrastructure of each elevator operator.

Therefore, this infrastructure contains the software modules that allow the information coming from the elevators to be processed and treated with the aim to:

• collect the information coming from the elevators
• process the information of each elevator and adapt it to a defined format (based on JSON) for its subsequent sending to the API in charge of processing the information
• anonymise the information collected from the elevators
• provide secure (for example certificate-based) communications with the API
• manage the receipt of alerts from the data-processing module

At the component level, it must include software modules with the purpose of performing lift information processing functions to adapt it to a standardised format based on a JSON (the data in that format will be stored and analyzed, but it is important to have a common format for the information that is stored in the data processing module), such a JSON-based exchange format.

Data Processing Infrastructure (Red Box)

At the component level, it must include at least:

• Rest API client access interface: A Rest API interface will be deployed. This interface will be responsible for receiving the requests from the clients, transmitting the documents to the Event Manager (through the Event Server to compile and consistency check process the data, as well as through Data Ingest Manager to publish the data to the Event Manager), as well as providing the response to these requests generated by the python modules.
• Event Manager: Acting as an interface to the DataLake for data ingestion will be done through an interface (for example based on Rest API) to manage the data that is uploaded to the DataLake.
• DataLake: Data Repository where three databases will be used to organize the data in three levels:
  • DB1: A non-relational database (mongoDB) will be used to host the raw data in order to host the data from the elevator operators’ infrastructure in the DataLake. The fact that the Database used is of a non-relational type is due to the fact that this characteristic provides the Database with the necessary flexibility to manage records with a flexible structure. The latter is necessary since it is expected that different elevator installations have different data.
  • DB2: PostgreSQL will be used as the relational database. This type of database allows less flexibility (predefined fields) than non-relational ones but a faster response time. This last consideration is important since this database stores the results of the processing such as the detected alerts. The python modules that respond to the Rest API interface read the data stored in this database.
  • DB3: PostgreSQL will be used as the relational database. It will contain the useful data to later apply ML or generate the data visualisation.

• Data standardisation and structuring module in order to perform elevator information processing to adapt it to a standardized format
• Data analytics module and ML core in which all the necessary tools are deployed to carry out the processing and structuring of data, data analysis and generation of relevant metrics for consultation in this module and based on the analytics the generation of alerts is carried out
• Visualisation module that entails the generation of a dashboard as a summary of information visualising, for example, relevant metrics, incidents/anomalies detected
• Module for managing alerts: Based on the alerts detected, a communication will be made with the elevator operator’s infrastructure to inform them of the alerts detected.

In order to include new data processing algorithms, the proposed system will be carried out in a scalable way that allows new algorithms to be integrated into the different components of the platform in order to offer new services from different providers in the future.

6. Operating Scheme

According to the proposed technical approach, a user of the platform — an elevator maintenance service company — must have a previously developed infrastructure for the storage of data from their connected elevators.

These data, or a subset of them corresponding to a series of elevators defined by the elevator operator, must be normalised and standardised to the format defined by the platform before being entered into it.

The transfer of data to the analysis platform will be carried out through open APIs that will allow the information to be anonymised and transferred to the platform.

Also through APIs, the elevator operator will receive valuable information from the application of the analysis algorithms available on the platform and selected by the user according to their interest.

In any case, the platform will offer initial information to the user through a connector with a BI tool, resulting from the algorithms considered in the core of the platform (Data analytics module and ML core) in addition to the alerts generated by the platform.

7. Management Scheme

One of the objectives of the project is the creation of a platform to provide services to different companies in the lift industry. To accomplish this objective, the results of the project must meet technical and economic feasibility criteria, including:

• Technical feasibility: The platform must yield results that allow further improvements to its clients operating scheme. For this, it must collect a sufficient volume of data to provide quality results.
• Economic feasibility: The end-to-end process must be profitable for all participants and introduce efficiencies that improve the competitiveness for its clients and the industry in general.

The viability of the project will be positive as long as it is positive for each of the participants:

• Elevator maintenance companies: The investment to connect to the platform plus the recurring payment for its use must be lower than the efficiencies obtained by using the platform.
• Technology companies: The investment to connect to the platform must be less than the expected margin due to the increase in technology sales.
• Algorithm companies: The investment to develop and maintain an algorithm must be less than the income you will get from using it.
• Platform: The investment to be made to create the platform and the operating costs must be less than the income to be obtained from the clients, extracting the payments to the algorithm companies.
• End customer: The benefits obtained in terms of availability and quality of service must exceed the costs of the solution.

References

[1] Lasi, H., Fettke, P., Kemper, HG. et al., Industry 4.0, Springer, Business & information systems engineering 6, 239–242 (2014). doi.org/10.1007/s12599-014-0334-4
[2] Atzori, L.; Iera, A.; Morabito; G. The Internet of Things: A survey. Computer Networks, Vol. 54, No. 15, octubre de 2010, pp. 2787-2805.
[3] Madakam, Somayya, et al. Internet of Things (IoT): A literature review. Journal of Computer and Communications, 2015, vol. 3, no 05, p. 164.
[4] Al-Abassi, Abdulrahman, et al., Industrial big data analytics: challenges and opportunities. Handbook of big data privacy. Springer, Cham, 2020. 37-61.
[5] Bishop, Christopher M., and Nasser M. Nasrabadi. Pattern recognition and machine learning. Vol. 4. No. 4. New York: Springer, 2006.
[6] Russell, Stuart J. Artificial intelligence a modern approach. Pearson Education, Inc., 2010.
[7] Goodfellow, Ian, Yoshua Bengio, and Aaron Courville. Deep learning. MIT press, 2016.
[8] Zhang, S., Wang, B., & Habetler, T. G. (2019). Machine learning and deep learning algorithms for bearing fault diagnostics-a comprehensive review. arXiv preprint arXiv:1901.08247
[9] S. Chai et al., A Non-Intrusive Deep Learning Based Diagnosis System for Elevators, in IEEE Access, vol. 9, pp. 20993-21003, 2021, doi: 10.1109/ACCESS.2021.3053858.
[10] Commercial references: TKE. MAX. tkelevator.com/es-es/productos/max/ (Oct. 2022) / Schindler AHEAD. us.schindler.com/en/services/digital.html (Oct. 2022) / KONE. kone.com/en/products-and-services/maintenance-and-modernization/24-7-connected-services.aspx / Otis. otis.com/en/uk/products-services/otis-signatureservice/otis-one