Low-OPEX Edge Controllers for Automated Sand Management

← Back to Solutions

The Problem

In oil & gas production, sand is an inevitable byproduct. Wells produce sand-laden fluids that must be separated before the clean stream moves downstream. Sand accumulates inside separators and is periodically purged by opening a sequence of dump valves to flush sand-laden water into a collection tank.

Traditionally, this dump cycle is triggered manually — requiring field personnel to be physically present at the separator, in close proximity to high-pressure equipment and hazardous fluids. This creates two problems:

• Safety risk: operators working near pressurized vessels and sand-laden discharge lines are exposed to serious HSE hazards
• Inefficiency: the dump sequence itself is straightforward — open valves in order on a preset schedule — yet it demands a person on site every time

Automation is the obvious answer, but existing PLC-based solutions come with a heavy price tag — expensive hardware, proprietary HMI panels, and integration costs that are hard to justify for what is fundamentally a simple sequencing task.

The System

A typical sand management loop consists of a separator vessel, a series of gate valves, a choke for flow restriction, and an open collection tank. A level transducer on the separator provides feedback to the controller, which orchestrates the valve sequence automatically. Where tank-state awareness is needed, an optional fixed camera can run a vision model on its own onboard chip and report a full or otherwise concerning tank condition.

FloLink's Approach

Instead of a traditional PLC + HMI stack, we developed custom firmware running on a low-cost edge controller. The controller handles the entire dump valve sequence — reading the level transducer, activating valves in the correct order, and managing timing — all without a PLC in the loop.

Custom Firmware

• Purpose-built for the sand dump sequence — no bloated general-purpose PLC runtime
• Reads separator level data and triggers automated dump cycles on a configurable schedule
• Activates the four gate valves and choke in the correct sequence with user-defined timing intervals
• Built-in safety interlocks and fail-safe defaults

Optional: On-Camera Vision for Tank-State Awareness

A fixed camera can be aimed at the dump tank, with its onboard chip running a site-trained computer vision model at the edge. The model can identify when the tank is full, as well as other operator-defined visual states of concern, without requiring video to be continuously watched by personnel.

When it recognizes a configured condition, the vision system can send a tank-state event to the edge controller so its configured control and safety logic can respond. It can also notify assigned personnel directly by text message or phone call, independently of the controller.

• On-camera processing keeps visual analysis at the tank and avoids a continuous video-stream dependency
• Site-trained model recognizes a full tank and other operator-defined conditions of concern
• Configurable controller events support the appropriate response within the existing safety logic
• Direct SMS or phone-call alerts bring the right person in only when attention is needed

Mobile-First Interface — No HMI Needed

One of the biggest advantages of an edge controller over a PLC is native connectivity. The controller hosts its own web interface accessible from any phone, tablet, or desktop browser — no proprietary HMI panel, no extra hardware, no license fees.

• Monitor system status, valve states, and separator level in real time
• Adjust dump schedules and timing parameters remotely
• Receive alerts and notifications without being on location
• Intuitive UI — no PLC programming knowledge required to operate

Optional: Cloud-Based ML for Schedule Optimization

Because we are an ML company at heart, the low-OPEX edge controller can optionally enable the operator to use ML models in the cloud to adjust dump schedule parameters as operational conditions change.

The ML model stack sits independent of the critical control hardware, entirely in the cloud, and can be easily turned off or overwritten by the operator at any time. It doesn't claim to be smarter than the human — it is simply there to assist when the operator wants it, or when they are busy with other tasks.

• ML recommendations are advisory only — the operator always has final control
• Cloud-based model is fully decoupled from on-site control logic
• Can be disabled with a single toggle — no impact on core automation
• Adapts dump scheduling suggestions as well conditions and sand loading change over time

Key Results