How does the asset management system structure be designed? Layers and integration points are analysed.
RFID Knowledge

How does the asset management system structure be designed? Layers and integration points are analysed.

PublishedShanghai Zejie Smart Technology Co., Ltd.

Last year we worked with a precision components plant with annual output worth more than CNY 100 million. It had more than 200 machines and over 600 assets once forklifts, computers, and test instruments were included. Records were spread across the equipment department's Excel sheets, IT's office automation system, and finance's ERP module. At year-end, the three sets of records did not match, and discrepancies with physical assets exceeded 8%. The problem was not that employees were careless. The system architecture was fragmented, with no shared asset master data or clearly defined integration responsibilities.

The core of the asset management system architecture design is:Scattered, border-defined and integrated. Usually, three layers of the responsibilities — access, business, data levels — are broken down and then connected to the ERP, MES, IOT platforms through a single interface gateway that builds on the full life cycle management of assets from procurement, accounting and transport to end-of-life.

Background: Why is the monosystem not holding?

Many companies put all their functions into one system early on: purchase registers, depreciation calculations, maintenance worksheets, inventory processes running in a large module. It's a little more economical when assets are small, and once they're over a thousand, there's more than five types of equipment, the problem is concentrated.

When we combe an old system of a logistics enterprise, we see that the inventory function is reset by a field, and the entire application is restarted; when Iot data is accessed, because the alignment is too tight, and the scanned worksheet module follows Cardon. The mono-structure is very laborious in terms of functional iteratives and data extensions. And the more difficult thing is to integrate -- the other side would have wanted the system to connect to the financial software, but the finance and business modules would share a database table, and any upgrade could have made the other side wrong.

Layering: detaching duties to three levels

A mature asset management system typically uses a three-tiered structure, each of which is self-directed and crosses the standard interface.

Access level: unified acquisition entrance

This layer is responsible for bringing in data sources - bar-code scanners, RFID readers, IOT sensors, ERP master data, mobile end APP. The access layer does not process business logic, but only protocol conversion and data integration.

For example, the same plant has both old RS485 sensors and new MQTT devices, which are converted to a single message format for the interface and sent to the business layer for processing. Otherwise, the later maintenance costs will run out of control if more than a dozen protocol adapters are maintained at the same time.

Business level: Core rules here.

Asset master data management, depreciation rules, maintenance worksheet flow, spare parts inventory logic, authority approval - all business process-related elements are concentrated on this level. Common practice is to use microservices or modular design and split by business domain (financial domain, transport domain, inventory area).

For example, the average age method, double-deduction of balances and workload method are used as a depreciation to encapsulate as independent services, with the transfer of assets from different subsidiaries and countries to one set of interfaces without redevelopment.

Data layer: main data separated from time series data

The data layer answers two questions: which assets are now, and what is the historical trajectory. The former uses relational data to store master data, while the latter records operational parameters collected by sensors in a special time series database (InfluxDB, TDengine etc.).

When Shanghai selects these options, it is common practice to place operational parameters such as car-mount humidity in a single time series, with the main data being placed on MySQL. The two sides link by asset numbering, aggregate by time window when searching, and not drag each other, and allow space for subsequent predictive maintenance.

Integrated elements: which borders must be secured

The structure is good, and it's all tested at the integration. Three of the integration points are key.

Sync with ERP's master data

The asset card is to be aligned between the asset management system and ERP. It is recommended that one-way synchronization should be used, with change notifications supported by: new assets, changes in status, transfers, end-of-life from the asset system, and only ERP ends receive and update. This avoids the two-way sync-generated main data drift — covering each other, and nobody knows which record is true. On a synchronized frequency, master data triggers an event, without taking time-trip inquiries.

Data flow back with IOT platform

The IoT platform is responsible for equipment networking and real-time data acquisition, the asset system is responsible for matching collected data to specific assets and triggers subsequent business rules (e.g., automatic inspection of a surveyer's report at continuous high temperatures). The integrated approach recommends the use of message queues (Kafka, RabbitMQ), where anaesthesia processes avoids each other from getting stuck in a peak hour system.

A weak interface with OA, HR

The approval process, the personnel organization structure is usually maintained in OA or HR systems. Asset systems simply need to call their interfaces to access organizational trees and people information, so do not write approval logic into their own databases - change a process to an integrator.

A few pits that can easily be stepped on during the operation.

A better structure, if not implemented, can also deform. The following three pits are most frequent.

Pit one: The layer is divided into two skins.The surface is layered, and in fact the data layer and the operating layer structure are combined, moving the database is equivalent to rewrite. The response is to strictly limit cross-layer calls -- the business layer can access data only through the DAO interface provided by the data layer, which is considered a breach code submission.

Pitfall two: Too many integrations turn into an unmanageable tangle.Each business system has its own interface, and dozens of interfaces are unclear. It is recommended that a single API gateway be introduced, all external interfaces registered at the level of the gateway, who calls, who is responsible for it, and how many times they are called have logs to check.

Pipe III: Time series data master data storage.The time series data must be stored in a separate place and accompanied by the data down sampling policy.

Landing recommendations: from which step

If the size of assets is less than 500 and not more than three types, first run through processes with a mature asset management system standard product without having to develop a stratification architecture from the outset.

When assets are over-thousands, of more than five types and need to be integrated with at least two external systems, it is recommended that they be redesigned in a stratification approach. Priority is given to regularizing primary data, consolidating access layers and gradually disaggregating the business layer by domain.

At any scale, one thing must be settled from the outset:Who defines the master data, what processes to change. The structure is a container, and the contents of the container are regulated so that the system can run in a stable manner over time.