What are the ways in which an automated inventory of fixed assets is conducted?
RFID Knowledge

What are the ways in which an automated inventory of fixed assets is conducted?

PublishedShanghai Zejie Smart Technology Co., Ltd.

The automatic inventory of fixed assets is mainly five ways, in order of degree of automation, from a low to high: manual inventory, bar/two-dimensional scan, RFID radio frequency recognition, A visual identification, and an AI-networking+ location technology. The vision of RFID and AI are two main paths for the current enterprise to upgrade the inventory, which are suitable for large asset numbers, relatively concentrated scenarios, and more flexible, with lower deployment thresholds.

Point 1: Manual inventory - minimal automation, but still the starting point for many enterprises

It is entirely by the person holding a paper or Excel form, and an asset is checked against the number, labeled, signed, summarized.

At a midsize plant with 10,000 assets, one inventory cycle usually takes one to two months and needs at least three to five people. The main sources of error are missed items, incorrect counts, and detached tags. Industry experience puts the error rate for this method at 3%–5%.

The advantage is zero hardware input, the disadvantage being that data cannot be synchronized in real time, results lag behind and accountability is difficult to trace. This approach is appropriate for a small number of assets (within hundreds), fixed positions, low frequency of change or as a transitional option for the new system.

Element 2: Bar/D-Scan count - the most common path to upgrading for SMEs

Attach a barcode or QR code label to each asset. Inventory staff go to the asset and scan it with a PDA, an industrial handheld with a built-in scanner, or a mobile app. The data is written to the back end automatically.

This approach is semi-automated and achieves over 99% accuracy. As long as tags are present and unobstructed, incorrect reads are rare. A count of 10,000 assets drops from months to three to five days and can be done by two or three people.

The drawbacks are also clear: someone must be physically present and aim the scanner at the tag from close range. If assets are high on a shelf, in an underground utility shaft, or covered with heavy oil and cannot be scanned, someone has to come back. Tags also have a limited life. Thermal-printed barcodes fade after two to three years. Use PET, a more wear-resistant plastic, for outdoor use or hot workshops.

If your records cover a few thousand to 10,000 or 20,000 assets and they are not too dispersed, barcodes offer the best value. For these customers, Shanghai Zejie Technology generally helps establish clear asset coding rules before launching scan-based stocktaking. That makes subsequent integration with the finance module much easier.

Point 3: RFID radio frequency recognition (RFI) inventory - good choice for mass, high frequency sub-counting

RFID is a single electronic label for each asset, and the inventory operator does not have to go to the asset, but whenever he walks in a certain range with a reader, the tag information is automatically read in bulk. The effective reading distance of UHF RFID is usually between 3 and 10 metres, and a cart-type reader can read hundreds of labels at once.

Automation is notably greater than with barcodes. Two people can count 10,000 assets in one to two days at 95%–99% accuracy. Metal or liquid environments lower read rates and call for on-metal tags.

The RFID's greatest advantage is "non-contact, mass reading", suitable for warehouse shelves, tool cabinets, equipment-intensive workshops. The limits are clear: labels are worth 0.5 to 3 dollars, one order of magnitude higher than bar codes; metals, liquids, dense stacking can jam signals; readers themselves cost thousands to tens of thousands of dollars.

Before choosing RFID, calculate two things: the upfront cost of tags and hardware, and the labor hours saved. Businesses with more than 5,000 assets and more than two counts a year can usually recover their investment in about two years.

Element 4: AI Visual Identification Inventory - the fastest-growing programme in the last two years

No labels, several cameras on the shelf or in the area, the AID model automatically recognizes the appearance of assets, their location and number, data directly into the system. The next version also makes an asset status judgement, such as identifying device to indicate light colours, tab readings.

This way, inventory counting becomes "active" to "self-run" and automating close to the RFID, but more flexible: no asset needs to be attached to each asset, no move after the camera is installed. Identification of accuracy and environmentalally relevant, clean indoor scenes up to 90% or more, light differentials, and a lot of covert scenes will fall to 70 to 80%.

The cost structure is different: hardware is a camera and edge computing device (a small server that is placed on site to process the scene, software is trained in AI models, with higher input than RFID, but more time-consuming. Fits to asset appearances are very diverse, not easy to label (high temperature, corrosive environment), or a scenario that requires state monitoring.

Two pits need to be noted: one is the blind spot of the camera, the back of the shelf, the assets in the cabinet are not visible; and the other is that the AID model needs to be optimized, and changes in the appearance of assets, and people's cover can affect recognition. The project experience with Shanghai-based Selectory Technologies is that it takes time to do a field survey before we can get this camera nodding and re-lighting program.

Element 5: Networking of goods + positioning technology - the highest form of real-time online assets

Loading a locator module on an asset (UWB, Bluetooth AoA, BeiDou, etc.) provides real-time access to the location of each asset, state and moving trajectory. The count is no longer a "regular" move, but it's the latest data at any time that you open the system.

This is the most automated way, but it's also the most costly and complex way. The accuracy of positioning varies from a rice to an ami scale, with prices very different: AoA models in Bluetooth each for dozens of dollars, UWB modules for hundreds of dollars each. Systems build involves base station deployment, signal cover, platform interface, million-scale start.

These programs are currently used in three types of scenarios: real-time monitoring of precious equipment and instrumentation, management of vehicles at large sites and mines, tracking of high-value consumables in hospitals and laboratories. The conventional fixed assets of a general enterprise do not recommend going to this level, and the value for money is inappropriate.

Element 6: How to judge the degree of automation when choosing a model

It's enough to look at four dimensions:

When you actually select a type, you start with the number of assets, location distribution, frequency count, single asset value, and then you score against these four dimensions, the solution is pretty clear. Budget is tight, fewer assets start from bar code, more assets, high-frequency counts are directly RFID or AI visualized, only a few scenarios that are valuable assets and must be tracked in real time can only consider IOT.