Proposed Concept · Engineering R&D

One IoT brain.
Multiple drying units.

Collapsible by design. Controlled by IoT. Monitored by Agentic AI.

A modular drying platform combining a collapsible aluminium structure, hanging rack system, food-grade heat-resistant enclosure, IoT-based environmental control and Agentic AI anomaly monitoring.

Proof of Concept Initial application: fucuk drying Preliminary design To be validated

Proposed engineering concept. No performance results, certifications or commercial validation are claimed. Dimensions and costs are preliminary estimates only.

COLLAPSED AIR IN AIR OUT BLOWER HEATER SoC RASPBERRY PI 5 EDGE / IoT GATEWAY TEMP · RH AIRFLOW LOAD CELL AGENTIC AI ANOMALY MONITOR STATUS NOMINAL 1800 mm APPROX IoT · continuous control AI · anomaly monitor SAFETY · independent cutoff
One IoT brain. Multiple drying units.

Add drying capacity without rebuilding the intelligence layer.

Recurring design principle
01 · The Approach

A different approach to drying

Conventional drying relies on fixed racks, flat trays and manual supervision. This proposed concept combines three things that are usually separate: a physical structure that can collapse and scale, a sensing layer that measures the drying environment continuously, and an intelligence layer that watches for abnormal behaviour across the whole process.

01 · MODULAR PHYSICAL

Collapsible

A physical drying unit that can fold, move, store and scale — designed for assembly, disassembly, transport and storage rather than a permanent fixed installation.

02 · SENSING LAYER

Connected

IoT sensors continuously measure the drying environment — temperature, relative humidity, airflow, product weight, duration and ambient conditions — giving the system a live picture of the process.

03 · INTELLIGENCE LAYER

Intelligent

Agentic AI monitors the entire process, analyses relationships between sensors and operating conditions, and looks for abnormal behaviour rather than simply displaying readings.

Collapsible drying unit STRUCTURE · HANGING RACK ENCLOSURE · AIRFLOW IoT sensing & control TEMP · RH · AIRFLOW · WEIGHT BLOWER · HEATER · CYCLE Agentic AI monitoring OBSERVE · ANALYSE · DETECT REASON · ADVISORY ACTION sensor feedback · continuous loop INDEPENDENT SAFETY LAYER (SEPARATE)
Physical modularity+Connected sensing+Intelligent monitoring
02 · The Drying Unit

Designed to collapse. Designed to scale.

The proposed unit is built around a collapsible aluminium frame with a removable hanging rack system, wrapped in a flexible enclosure. Every major assembly is intended to be folded, removed or added without rebuilding the whole system.

① COLLAPSIBLE ALUMINIUM FRAME ④ FOOD-GRADE HEAT-RESISTANT FLEXIBLE ENCLOSURE (PROPOSED) ⑤ AIR INLET ⑥ AIR OUTLET ③ HANGING RAILS (2–3, ADJUSTABLE) ② COLLAPSIBLE HANGING RACK (REMOVABLE / FOLDABLE) FUCUK HUNG VERTICALLY ⑦ BLOWER ⑧ SENSORS ⑨ HEATER ⑩ CONTROL / ELECTRICAL MODULE RASPBERRY PI · SSR · POWER · SAFETY
1Collapsible aluminium frame — foldable posts, beams and bracing
2Collapsible hanging rack — removable, foldable, modular
3Hanging rails — 2–3 adjustable horizontal rails
4Flexible enclosure — proposed food-grade heat-resistant material
5Air inlet — controlled intake
6Air outlet — controlled exhaust
7Blower — adjustable airflow
8Sensors — temperature, RH, airflow, weight
9Heater — adjustable heat input
10Control / electrical module — edge controller, actuators, power, safety

Why a hanging rack instead of flat trays

Conventional fucuk drying commonly relies on flat trays. This concept instead proposes hanging the product vertically from a removable rack, so the drying surface is exposed on both sides and the rack itself can be lifted out, folded and stored. Whether hanging improves drying uniformity is a core question this PoC is designed to answer.

Modular by intent

The rack, rails and enclosure are treated as separate modules rather than parts of one welded box. That is what allows the unit to be disassembled for transport, stored flat, and expanded by adding more units to the same control system.

Proposed Concept — assembly, hinge geometry, rack load rating and enclosure attachment method are preliminary and require engineering review and prototype validation.

03 · Geometry

Preliminary prototype geometry

The starting geometry below is a first-pass envelope for a single collapsible unit. It exists to give the prototype a concrete size to engineer against — not as a fixed final specification.

1200 mm W 1800 mm H 800 mm D PRELIMINARY ENVELOPE — NOT FINAL GEOMETRY
Preliminary envelope 1,200 × 800 × 1,800 mm
Estimated chamber volume ±1.5–1.7 m³

Proposed hanging rack

  • 2–3 horizontal hanging rails
  • Approximately 700–1,000 mm rail length
  • Adjustable hanging spacing
  • Removable / foldable supports

Preliminary engineering dimensions. Final geometry, hanging capacity and airflow configuration require prototype validation.

04 · Material System

Lightweight structure. Flexible drying environment.

The material system is deliberately split into layers so each can be specified, sourced and validated independently. Material selection must be validated against actual operating temperature, durability, cleaning requirements and food-contact requirements.

Structure
25–30 mm aluminium profile. Chosen for a favourable strength-to-weight ratio, corrosion resistance and the ability to use standard modular connectors — which is what makes a collapsible frame practical to build and rebuild.
Inner enclosure
Proposed food-grade, heat-resistant flexible material. A flexible skin keeps the unit lightweight and foldable. The exact material grade and its suitability for continuous food-contact operating conditions remain to be validated.
Hanging system
Food-contact suitable aluminium / stainless components. Rails, hooks and supports are intended to be cleanable and compatible with the drying environment; specific grade selection requires validation for the actual process.
Airflow
Appropriate ducting, inlet and outlet components. Ducting is sized for the PoC airflow target and positioned to give even distribution; final configuration depends on measured airflow testing.
FRAME · ENCLOSURE · HANGING SYSTEM · AIRFLOW LAYERS

What must be validated

Material compatibility with the actual operating temperature, cleaning and sanitation routine, durability under repeated folding, and food-contact suitability. No food-contact certification is claimed for any proposed material.

Why layer the materials

Separating structure, enclosure, hanging system and airflow lets the team validate each layer on its own — and swap one material without redesigning the entire unit.

05 · PoC Hardware

Prototype equipment

The basic PoC bill of materials below covers the sensing, control, actuation and safety hardware needed to build one functional drying unit and start generating real operating data. Sensor classes and power figures are initial targets, not final selections.

01

Raspberry Pi 5

Edge controller · IoT gateway

Central edge computer that reads sensors, runs the deterministic control logic and hosts the data/control engine. Also the bridge to the Agentic AI layer.

02

Temperature sensor

DS18B20 or RTD / PT100 class

Measures chamber and airflow temperature. Sensor class and placement are preliminary; accuracy and response time require validation.

03

Humidity sensor

SHT31 / SHT35 class

Measures relative humidity inside the drying environment — a primary indicator of how the drying process is progressing.

04

Airflow sensor

Air velocity / airflow measurement

Measures air movement through the chamber. Intended to confirm the blower is producing the airflow the control layer expects.

05

Load cell + HX711

Product weight / drying progress

Tracks product weight over time as an indicator of moisture loss, giving the AI layer a drying-progress signal rather than only environment data.

06

Blower

Initial PoC target: adjustable 100–300 CFM

Moves air through the chamber. Adjustable speed lets the control layer vary airflow as part of the drying cycle. Target figure is preliminary.

07

Heater

Initial PoC target: adjustable 1–2 kW

Supplies controlled heat within the enclosure. Power target is a starting point for the PoC and depends on chamber volume and insulation performance.

08

SSR / relay / controller

Actuator switching

Switches the blower and heater under command from the control layer. Chosen for reliable, repeatable actuation rather than manual control.

09

Power supply

24 V DC and required converters

Provides regulated DC power for sensors, controller and actuators, with the necessary converters for each component rail.

10

Safety cutoff

Independent thermal + electrical protection

A separate protection path — over-temperature cutout and electrical protection — that does not depend on the AI layer or the main control software.

11

Ducting

Controlled air inlet / outlet

Directs intake and exhaust so the airflow path can be measured and adjusted. Final layout depends on measured distribution testing.

12

Wiring & enclosure

Industrial-style protected electronics

Protects the electronics, routes cabling safely and keeps the control module serviceable without dismantling the drying unit.

Optional

Optional sensors

Ambient · differential · energy · door

Ambient temperature, ambient humidity, differential pressure, energy/current measurement and door/opening sensors. Included only where the PoC needs them.

Component classes, sensor accuracy, blower and heater sizing, and the safety protection arrangement are preliminary PoC targets and require engineering validation before procurement or build.

06 · System Architecture

From drying unit to intelligent loop

Sensors feed a central edge controller. The deterministic control engine runs the drying cycle, while Agentic AI sits alongside it as a monitoring and reasoning layer. Control outputs drive the blower and heater, and the resulting change is measured again by the sensors — closing the loop.

COLLAPSIBLE DRYING UNITS structure · rack · enclosure
SENSORS temp · RH · airflow · weight
CENTRAL IoT / RASPBERRY PI edge controller · gateway
DATA + CONTROL ENGINE deterministic rules · logging
AGENTIC AI observe · analyse · detect · reason
CONTROL / MONITORING commands · alerts · dashboard
BLOWER + HEATER + AIRFLOW actuation
DRYING PROCESS product drying in progress
↺ sensor feedback · continuous loop
Deterministic control and Agentic AI monitoring are separate layers running on shared data.
Dryers → Sensors → IoT / Pi → Agentic AI → Blower · Heater → Drying ↺
07 · Control Architecture

Three layers. One intelligent drying system.

The architecture deliberately separates what controls the dryer, what monitors it intelligently, and what protects it. Each layer has a distinct responsibility, and the safety layer is independent of both the control software and the AI.

Layer 01

Control layer

IoT-based deterministic control. Handles the routine drying cycle using fixed, predictable rules — this is the layer that normally keeps the system running.

Blower Heater Airflow Drying parameters
Layer 02

Intelligence layer

Agentic AI. Watches behaviour across sensors and operating conditions, looks for patterns that do not fit normal operation, reasons about possible causes, and interprets events for the operator.

Monitoring Pattern analysis Anomaly detection Reasoning Adaptive response Event interpretation
Layer 03

Safety layer

Independent protection. Operates separately from the control software and the AI layer, and takes priority for critical conditions.

Over-temperature Electrical protection Emergency shutdown Safe-state operation
AI IS NOT THE PRIMARY SAFETY MECHANISM.

The separation of control, intelligence and safety is a design intent for the PoC. Final safety architecture, protection ratings and fail-safe behaviour require engineering validation against applicable standards.

08 · Agentic AI

AI that watches the process — not just the dashboard

The AI layer is not a chart that a person has to read. It observes system behaviour continuously, analyses relationships between sensors and operating conditions, and looks for abnormal behaviour. Where authorised, it can assist with corrective action — but the deterministic control layer handles normal operation, and the safety layer always takes priority.

01

SENSE

Read continuous sensor data from every drying unit.

02

ANALYSE

Understand the current drying conditions as a whole, not one reading at a time.

03

DETECT

Identify patterns that deviate from expected behaviour.

04

REASON

Evaluate possible causes and relationships between sensors and operating conditions.

05

ACT

Where authorised, initiate an appropriate adjustment through the control layer.

06

VERIFY

Check whether the system responds as expected after the action.

07

ESCALATE

Alert the operator or initiate safe-state logic when required.

SENSE 01 ANALYSE 02 DETECT 03 REASON 04 ACT 05 VERIFY 06 ESCALATE 07 verify feeds back into the next observation cycle
WHAT IT IS NOT

Not the control layer

Normal drying operation is handled primarily by the deterministic IoT control layer. The AI observes and reasons — it does not replace routine control logic.

WHAT IT IS NOT

Not the safety layer

Critical protection and emergency shutdown remain the responsibility of an independent safety layer that does not depend on the AI.

WHAT IT IS

An anomaly watch layer

A continuous intelligence layer that watches the process for abnormal behaviour, interprets it, and supports the operator with analysis and, where authorised, corrective action.

09 · Operating States

Designed for normal operation. Built to notice abnormality.

The system is expected to spend almost all of its time in normal operation. The value of a separate intelligence layer is that it can recognise the early signs of abnormality before they become a serious condition — and hand the serious cases to independent safety controls.

Layer 01 · Control

Normal

The IoT system automatically maintains configured drying conditions. No intervention required.

Layer 01 + 02

Minor deviation

The system observes the drift and may adjust operating parameters within its normal control range.

Layer 02 · Intelligence

Abnormal pattern

Agentic AI identifies unexpected behaviour, investigates relationships between signals and raises an interpreted alert.

Layer 03 · Safety

Serious abnormality

Independent safety controls take priority and the system can enter a defined safe state.

NORMAL control layer MINOR control + observe ABNORMAL AI investigates SERIOUS safety takes over SAFETY PRIORITY PROPOSED STATE MODEL — THRESHOLDS TO BE DEFINED DURING PoC

The four states describe an intended operating model. The thresholds, escalation rules and safe-state definition are preliminary and must be specified and validated during prototype development.

10 · Example Anomalies

What the intelligence layer is designed to look for

The examples below illustrate the kind of abnormal behaviour the proposed Agentic AI layer is intended to reason about. They describe a proposed function and PoC target — not a validated detection capability.

PROPOSED AI FUNCTION · PoC TARGET

Airflow anomaly

Sensor data

Blower is commanded on and draws expected current, but measured air velocity through the chamber unexpectedly decreases over time.

AI detection

AI correlates blower command, current draw and measured airflow, and flags airflow that no longer tracks the command.

Analysis

Possible causes considered: duct obstruction, filter loading, leakage in the enclosure, sensor drift or blower degradation.

Alert / action

Raise an interpreted alert with likely causes, and where authorised, adjust blower output or pause the cycle pending inspection.

PROPOSED AI FUNCTION · PoC TARGET

Temperature anomaly

Sensor data

Temperature behaviour does not correspond with the heater command — for example it keeps rising after the heater is switched off.

AI detection

AI compares commanded heat input against the temperature response curve and the expected rate of change.

Analysis

Possible causes considered: SSR stuck closed, sensor placement issue, insulation behaviour, or residual heat effects.

Alert / action

Escalate to the operator immediately; the independent safety layer can enforce an over-temperature cutoff.

PROPOSED AI FUNCTION · PoC TARGET

Humidity anomaly

Sensor data

Humidity remains unusually high despite operating conditions that would normally reduce it.

AI detection

AI tracks the humidity profile against airflow and temperature and flags a drying environment that is not progressing as expected.

Analysis

Possible causes considered: insufficient exhaust, recirculation of moist air, wet batch loading or a blocked outlet.

Alert / action

Recommend an airflow or cycle adjustment and notify the operator if the humidity profile does not recover.

PROPOSED AI FUNCTION · PoC TARGET

Weight anomaly

Sensor data

The product weight-loss curve deviates from the expected drying profile — too fast, too slow, or stalled.

AI detection

AI compares the measured weight-loss curve against the expected shape rather than a single threshold value.

Analysis

Possible causes considered: incorrect loading, uneven airflow distribution, load cell drift or a genuine process change.

Alert / action

Flag the batch for review, propose a drying-time adjustment, and record the deviation for later validation.

PROPOSED AI FUNCTION · PoC TARGET

Sensor anomaly

Sensor data

One sensor produces readings inconsistent with surrounding sensors measuring the same environment.

AI detection

AI compares redundant and related sensors to identify readings that do not fit the pattern of its neighbours.

Analysis

Possible causes considered: sensor failure, wiring fault, calibration drift or a localised physical condition.

Alert / action

Raise a sensor-health warning and, where redundant sensing exists, continue on the remaining sensors where safe to do so.

All anomaly examples above are proposed AI functions and PoC targets. Detection accuracy, false-positive behaviour and the correct response for each case must be established through prototype data, not assumed. Serious abnormalities are handled by the independent safety layer, not by the AI.

11 · Multi-Unit Scalability

Scale the drying capacity — not the intelligence infrastructure

To increase drying capacity, additional drying units can be added with their own sensors and drying hardware, while the central IoT and Agentic AI infrastructure can remain shared. The intelligence layer does not need to be rebuilt for every new unit.

CENTRAL IoT + AGENTIC AI ONE CONTROL + MONITORING LAYER DRYER 01 Sensors · Blower · Heater DRYER 02 Sensors · Blower · Heater DRYER 03 Sensors · Blower · Heater DRYER N Sensors · Blower · Heater …
1 UNIT → 2 UNITS → 4 UNITS → 8 UNITS → MULTIPLE UNITS

What stays shared

The central IoT/AI layer, the dashboard, the data history and the anomaly-monitoring logic are designed to be shared across units. Adding a dryer adds sensors and drying hardware — not another copy of the intelligence stack.

What must be validated

Actual network capacity, controller capacity, communication architecture, sensor addressing and control-loop timing all require validation. This concept does not imply unlimited scalability.

The scaling pattern is a proposed architecture. Practical limits on the number of units per controller, communication reliability and control-loop timing must be established by testing, not assumed.

12 · Central Dashboard

One dashboard. Every drying unit.

A single control surface is intended to show every connected unit together — environment readings, drying progress, actuator status, the drying curve, alerts and the event log. The panel below is a design mockup using clearly labelled simulated placeholder values.

Drying Operations Console SIMULATED PLACEHOLDER DATA — NOT TEST RESULTS
Dryer 01 NORMAL
Temp38.4 °C
RH42 %
Airflow1.8 m/s
Weight4.21 kg
BlowerON
HeaterON
Dryer 02 DRYING
Temp41.2 °C
RH55 %
Airflow2.4 m/s
Weight3.87 kg
BlowerON
HeaterON
Dryer 03 MONITORING
Temp39.7 °C
RH48 %
Airflow2.1 m/s
Weight4.05 kg
BlowerON
HeaterON
Dryer 04 ANOMALY
Temp44.9 °C
RH61 %
Airflow0.9 m/s
Weight4.42 kg
BlowerCHECK
HeaterHOLD

Drying curve · weight-loss profile (simulated)

0h time 24h kg DRYER 04 · DEVIATION DRYER 01 · EXPECTED

Event log (simulated)

09:14Dryer 01 cycle started
09:22Dryer 02 heater on · setpoint steady
09:31Dryer 03 entering monitoring window
09:44Dryer 04 airflow lower than expected
09:45AI: pattern flagged · duct restriction suspected
09:46Operator notified · awaiting review
TempRHAirflowProduct weightDrying time Blower statusHeater statusAI statusEnergy consumption Historical dataAlerts

The dashboard above is a design mockup. All values, curves, statuses and log entries are simulated placeholders used to show the intended layout. They are not measured results from a prototype or test.

13 · Preliminary Budget

Preliminary prototype budget

An estimated starting budget for the hardware needed to build a single functional PoC unit. Figures are ranges because component selection and fabrication method are not yet fixed.

ComponentEstimated cost (MYR)
Collapsible aluminium frameRM800–1,500
Collapsible hanging rackRM300–700
Food-grade heat-resistant enclosureRM300–700
BlowerRM250–600
HeaterRM150–350
Raspberry Pi 5 + storageRM450–650
Temperature sensorsRM80–200
Humidity sensorsRM100–250
Airflow sensorRM150–400
Load cell + HX711RM100–250
SSR / relay / controllersRM100–250
Power / electrical componentsRM200–400
Safety protectionRM150–300
Wiring / connectors / enclosureRM150–300
Air duct / inlet / outletRM150–300
Miscellaneous fabricationRM300–600
Preliminary hardware PoC estimate RM4,000–RM8,000

Preliminary estimate only. Actual cost depends on engineering specifications, component selection, fabrication method and prototype requirements.

Excluded: Software development Excluded: Agentic AI development Excluded: Engineering labour Excluded: Certification Excluded: Laboratory testing Excluded: Commercial tooling
14 · Development Roadmap

From basic prototype to multi-unit system

The roadmap moves deliberately from the simplest possible proof — sensors and a controller — through the mechanical build, the controlled process, the intelligence layer and finally multi-unit scaling.

PHASE 01

Sensor + Raspberry Pi PoC

Bring up sensors and the edge controller, prove reliable data capture and basic actuation before any mechanical build.

Proposed sequence
PHASE 02

Collapsible Aluminium + Hanging Rack Prototype

Build and test the folding frame, hanging rack and enclosure attachment. Verify assembly, disassembly and storage.

Proposed sequence
PHASE 03

Controlled Airflow + IoT System

Add blower, heater and airflow control. Establish a repeatable, monitored drying cycle.

Proposed sequence
PHASE 04

Agentic AI Anomaly Monitoring

Introduce the intelligence layer, define normal behaviour and implement anomaly observation and reasoning.

Proposed sequence
PHASE 05

Fucuk Drying Validation

Run real fucuk drying trials and compare behaviour against expected profiles. Generate operating data.

Proposed sequence
PHASE 06

Multi-Unit Scaling

Connect multiple units to the shared IoT/AI layer and test scaling limits, communication and control timing.

Proposed sequence

Phases and sequencing are proposed and may change as findings emerge. Durations are intentionally not stated because they depend on budget, fabrication and validation results.

15 · Validation Plan

From concept to measured performance

The purpose of the PoC is to produce real operating data. Until those measurements exist, the concept remains a design hypothesis rather than a demonstrated result.

01 Drying time
02 Temperature stability
03 Humidity profile
04 Airflow distribution
05 Weight-loss curve
06 Energy consumption
07 Batch consistency
08 Repeatability
09 Operator intervention
10 Sensor reliability
11 AI anomaly detection performance
PROPOSED CONCEPT PRELIMINARY DESIGN PROTOTYPE BUILD CONTROLLED TEST MEASURED DATA DATA COLLECTION PRECEDES ANY PERFORMANCE CLAIM

Stated plainly

The PoC is intended to generate real operating data before performance claims are made. No drying-time reduction, energy saving, throughput or food-safety claim is asserted on this page. Any future claim should be traceable to measured results from the prototype.

16 · The Concept

A drying platform, not just a dryer.

The physical system can collapse. The drying capacity can scale. IoT provides continuous control and visibility. Agentic AI provides an intelligence layer that watches for abnormal behaviour while independent safety controls protect the system.

1 UNIT SCALE MULTIPLE UNITS ONE CENTRAL IoT + AGENTIC AI LAYER CONTROL · MONITORING · ANOMALY WATCH
One IoT brain. Multiple drying units.

The recurring principle behind the whole platform.

Proposed concept · PoC

Where this goes next

Fucuk drying is the initial proof-of-concept application because it has a clear product, a known drying requirement and a measurable output. The underlying platform — collapsible structure, modular hanging, IoT control and Agentic AI monitoring — is intended to be adaptable to other drying applications after validation. That adaptability is a hypothesis to be tested, not a claim.

ENGINEERING DISCLAIMER

Read this as an R&D concept, not a product

This page describes a proposed engineering concept and proof-of-concept direction. It is prepared for technical discussion and prototype development planning.

  • The drying system is not commercially validated.
  • No specific drying-time reduction, energy saving, throughput or capacity claim is made.
  • No food-contact certification is claimed for any proposed material.
  • All dashboard values, curves and log entries shown are simulated placeholders.
  • Dimensions, hardware specifications and costs are preliminary estimates.
  • IoT control, Agentic AI and independent safety are separate functions; the safety layer is not dependent on AI.
  • Fucuk is the initial PoC application. Adaptability to other drying applications is a hypothesis, not a validated capability.

Proposed Concept · PoC · Preliminary Design · Estimated Cost · To Be Validated