How to build an asset management system based on RFID? Hardware selection and data acquisition design
RFID Knowledge

How to build an asset management system based on RFID? Hardware selection and data acquisition design

PublishedShanghai Zejie Smart Technology Co., Ltd.

Reconstructing systems triggered by a single inventory

Last winter, a precision machining plant had its annual stocktake. The warehouse held more than 5,000 cutting tools, gauges, and fixtures, with metal items packed closely together. Three employees spent a full three days scanning and recording each item, yet more than 200 records still did not match. The manager later approachedShanghai Picking Up Tech Ltd.I want to go over the programme.

When you get to the scene, the problem is clear: metal environments are highly disruptive, assets are dense, manual can't be controlled. RFID is used for this kind of scene. In short, it's a digital ID for the items, and in short, a radio wave remote identification, asset management, hardware selection and collection logic must be designed in tandem, and unopening will be problematic. We'll break down a few key points that were validated in that project.

Hardware selection: not as expensive as possible, but as the scene selection configuration

Label: URF or HF, first look at the asset material

RFID labels are multi-frequency, and the most common of asset inventories is UHF (UHF, 860-960 MHz). It reads distance, cleans dozens of labels at a time, suitable for bulk count. But metal surfaces interfere with UHF signals, and metal assets must be labelled as resistant to metals. We used a hard label of ceramic base, placed on steel knife handles, to read distance stable at 4-6 meters; it was replaced by normal softer adhesive tags on plastic case equipment, which can be reduced to one tenth of the former.

Three parameters are also to be considered when selecting a label:

Reader and antenna: the density of the placement determines the rate of failure

The reader is the device to launch and receive radio frequency signals, and the antenna determines the shape and extent of the signal's coverage. The width of the warehouse channel is typically 2-3 metres, and we usually have a set of directed antennas on each side of the tunnel, with beams pointing towards the shelf, and the reader is mounted on top of the bridge. This standard channel has an width of about 1.5-2 metres, and the tray is almost unreadable.

A common pitfall is assuming more antennas are better. UHF signals can overlap and cause interference. On one project, we initially installed six sets of antennas. Testing showed that the two sides of the portal interfered with each other and read tags on the opposite pallet. We reduced the number to four sets, adjusted their orientation, and lowered power from 30 dBm to 27 dBm. The missed-read rate fell from 3% to below 0.3%.

Fixed or handheld, by course of operation

Readers are in fixed and handheld formats. Fixed-based devices are installed on a channel or at a working place, assets are automatically collected, suitable for access to the library, and these actions with fixed nodes are mounted up and down. Handhelds are suitable for inspection, looking for single-ware assets, such as a manager looking for a specific mass, walking around the shelf with a handheld machine, and locating them in seconds.

The mix we suggest is a fixed set of key nodes (to and from the warehouse door, shelf entrance) and daily handheld. A customer who makes car parts is configured in this way, reducing the time taken from two days to two hours.

Data acquisition design: trigger logic, filter rules, upload protocols

When: Event triggers more than time-bound inquiry

Many teams start by having readers scan around the clock, flooding the back end with duplicate data. Use event triggers instead. An infrared sensor at the door can start a read only when a pallet passes, for example, or tie data collection to an outbound order in the ERP (enterprise resource planning system) so readings start only when the order takes effect. That controls data volume at the source.

We divided the trigger logic into four categories in a storage project:

  1. Repository Trigger: The WMS (warehouse management system) generates a receiving order → the reader starts → tag IDs are captured → storage locations are linked
  2. Out of the library trigger: Dispatch order becomes effective → gate opens → reader captures data → records are checked against the order → an alert flags any discrepancy
  3. Move Trigger: Scan the new location code with the handheld → write it to the tag's User memory → update the location in the back end
  4. Accelerator trigger: Start a count on the handheld → enter inventory mode → capture all tags in range → generate a discrepancy report.

How to process duplicate data: Add Time Window and RSSI thresholds

The same label is read many times -- the tray stops at the door for two seconds, and the reader may have collected five to eight records. This is not needed in the backstage, filtering by middle. The common method is to add a window: the same label will be kept only one in 10 seconds; then the RSSI threshold (the signal strength indicator, the higher the label, the closer) will be added, and the most powerful record of the signal will be kept.

There's also a detail on conflict prevention. UHF reader generally supports the EPC Gen2 protocol (ISO 18000-63), which incorporates collision-proof algorithms that can read hundreds of labels at a time. But if more than 200 labels appear at the same time in a dense scene, reading times are significantly longer (from 0.5 seconds to more than 2 seconds). Our approach is to move Q values from readers to a reasonable range, along with group reading strategies.

How do you get the data backstage? The middle is not going to be saved.

Readers usually have a set of SDKs or intermediates that are used to decipher original label data, filter, link business fields and then push them through HTTP, MQTT, WebSocket (three common data transfer protocols). This layer is not looking at it, actually taking on most of the dirty work.

The middle is at least four things to do:

We do our own middle, we use MQTT protocols (lighter than HTTP, for HF small data packs), and backstage buffers with messages lined up like RabbitMQ, which are distributed to WMS and asset management systems. Readers and business systems untie, readers add or decrease, network shaking does not affect core business.

A few holes were stepped on.

The metal shelf ate the signal in half.

One time, we went to the field test, and the lab had a reading-writing distance of 8 metres, and then we put a metal shelf on the site, which was reduced to 3 metres. Because the metal shelf itself was a huge reflection, and the signal was reflected several times into a blind zone. And finally, we changed the label -- it was placed on a plastic partition at each point, not on a metal plate; and it was also marked with a metal reflector outside the reader antenna, focusing the signal in the shelf corridor.

Liquid assets completely failed the UFF.

RFID encounters liquids with another headache. Water absorbs ultra-high frequency signals, and the liquid-filled barrel assets can barely read them. This is either a HF tag (13.56 MHz, which has greater but shorter penetration), or a RFID hard label outside the liquid container, which keeps some distance from the liquid. In our program for a chemical company, the top of the liquid material drums is equipped with an emaciated plastic card button label, which is above 5 cm to read the distance to about 4 m.

The labels are covered and read, but the angle is right.

Some people think that they can read with a label, but not. The label antenna is very directional, reading the distance as far as it is from the reader's antenna, and it's going to drop sharply when it's above 45 degrees. The first thing we do is teach field operators how to look at the label in the direction -- the position of double-sided tape will have to be in the general direction of the label's antenna.

Once the data comes in: Business systems are valuable

Raw RFID data is of little use on its own. It needs to be integrated with the asset management system and sometimes with ERP. An equipment rental customer initially wanted only automated counting. We suggested linking RFID data to the asset register, maintenance records, and depreciation calculations. After the changes, each count not only showed which items were present but could automatically report overdue maintenance and assets due for retirement. Asset turnover efficiency rose by almost 30%.

Two common practices are used in integration:

SMEs generally choose to become less integrated, and large groups are more suitable for deeper integration, but the costs and workload have also multiplied.

From planning to getting online, it's recommended that we move forward.

If you're planning to introduce a RFID asset management system to the company, in that order:

  1. Take a first count of the status quo.: Number of assets, material, distribution, motion lines (mobile routes), undestructable without these data
  2. Do small-scale piloting: select a warehouse or type of asset, run through label selection, antenna setting, trigger logic
  3. Real scene measurement: Actual operating load detection rate, data delay, simultaneous distribution
  4. - I'm gonna take it back.: Pilot runs steady and replicates to other regions

Hardware typically accounts for 40–50% of system costs, with middleware and integration making up the other half. On a tight budget, start with key points such as receiving and dispatch doors, then add handhelds and other features in phases.

The RFID asset management is not a few readers, but it's done with a few labels. The hardware selection determines what data you can get, the data acquisition design determines whether they can be useful information, and two things are tied together to make the system really work. Shanghai has validated this in several industries, from metal processing to chemical storage, from equipment leasing to large-scale manufacturing, and can produce quantifiable results.