Hardware

As the core component of the Hardware Track in the iGEM Competition, this project focuses on developing a small-scale experimental device for real-time monitoring and automatic feeding of methanol. Designed for the cultivation process of Pichia pastoris in laboratories, the device enables real-time monitoring and automatic regulation of methanol concentration. Different from traditional liquid sensors— which are prone to microbial contamination, have short service lives, and high costs—this device innovatively detects the gas-phase methanol concentration above the fermenter, then infers the liquid-phase methanol concentration through a gas-liquid conversion model, realizing non-invasive continuous monitoring.

The device integrates a sensing system (centered on the GT-CX series catalytic combustion-type methanol gas sensor) and a control system (based on the ARM-series STM32G070 microcontroller with an added fuzzy control algorithm). The total budget is controlled within 1,600 RMB, and it supports application in 10L-scale fermenters. The project design complies with national standards such as GB15322.1-2019 Combustible Gas Detectors—Part 1: Point-type Combustible Gas Detectors for Industrial and Commercial Use to ensure safety and reliability.

Guided by the philosophy of "precision, environmental friendliness, convenience, and economy", the project makes full use of the gas-liquid equilibrium principle to avoid sensor corrosion and contamination issues caused by direct immersion in liquid. By detecting volatile gases, it achieves indirect calculation of liquid concentration, reducing hardware costs (gas sensors cost approximately 700 RMB vs. thousands of RMB for liquid sensors) and maintenance difficulty.

Meanwhile, the device emphasizes modular design, supporting disassembly and expansion (e.g., reserved PLC interface) for easy iterative optimization in laboratories. At the software level, a gas-liquid exchange model and fuzzy algorithm are integrated to improve feeding accuracy and response speed, facilitating efficient heterologous protein expression in Pichia pastoris. This philosophy not only addresses the pain points of real-time monitoring in microbial fermentation but also provides a low-cost solution for sustainable biomanufacturing.

This device is suitable for microbial fermentation processes at the laboratory and pilot scales, especially for methanol-induced expression systems (e.g., cultivation of Pichia pastoris for recombinant protein production). Typical scenarios include:

- Real-time monitoring of methanol concentration (threshold: 60 g/L) and automatic intermittent feeding (10 - 20 g/48 h) during the induction phase of Pichia pastoris in 10L fermenters to avoid toxic inhibition;

- Pre-industrial verification platforms, supporting multi-channel peristaltic pump expansion to 100L scale;

- Educational and research auxiliary tools, such as studying the impact of methanol feeding on Pichia pastoris growth rate/metabolic rate (e.g., experiments on optimal feeding speed).

In addition, it can be extended to monitor other volatile substrates, such as ethanol or acetone fermentation scenarios.

The device adopts the GT-CX series point-type combustible gas detector (produced by Jinan Anbang Instrument Co., Ltd.), with the core gas-sensing element being an imported catalytic combustion-type sensor. This sensor operates based on the platinum wire catalytic element principle: when methanol gas in the air comes into contact with a pre-installed catalyst (e.g., precious metals on an alumina carrier), an oxidation reaction occurs under heating (operating voltage: 5 V). The released heat causes a change in the platinum wire resistance (∆R ∝ gas concentration). The microcontroller (MCU) collects the resistance signal, which undergoes amplification, filtering, and analog-to-digital conversion (ADC) to output a digital concentration value (ppm).

The sensor has a response time of < 30 s, a repeatability error of < ±3%, and operates in an environment of −20◦C to +50◦C with humidity < 95% RH. It has an explosion-proof grade of Exd IIC T6 Gb and complies with the GB3836.1-2000 standard. Calibration is performed using standard gas (e.g., 1% vol methanol), and the zero point/range is adjusted via a remote control or serial port to ensure linear output (0–100% LEL).

Gas-liquid conversion is based on Henry’s Law, which describes the equilibrium of volatile solutes at the gas-liquid interface. Corrections are made for the physicochemical properties of the culture solution, mass transfer efficiency, and metabolic consumption according to the fermentation conditions of Pichia pastoris in the fermenter (see Section 4 for details).

Algorithm process: Sensor signal filtering (UAF42 low-pass filter) → ADC conversion (ADS1220) → control system calculation → signal conversion → output control signal.

A fuzzy algorithm is added to the microcontroller for optimization: the inputs are ∆C (concentration deviation) and dC/dt (rate of change), and the output is the feeding pulse width (PW). The rule base is based on the Mamdani model (e.g., if ∆C is large and dC/dt is positive, then PW is large) to achieve adaptive regulation.

Real-time collection of gas-phase methanol concentration, supporting temperature/pressure compensation.

The core component is the GT-CX series detector (size: ϕ140 x 120 mm, weight: 1.2 kg), which integrates a catalytic combustion probe and a buzzer alarm (threshold: 20% LEL). Auxiliary components include an ADS1220 analog-to-digital converter (24-bit precision) and a TI OPA2188 low-noise operational amplifier (gain: 10x).

The probe is placed in the gas phase above the fermenter, connected via a three-way interface, and the cable is routed out through a sealed hole (IP65 protection). A 4 - 20 mA analog output and RS485 digital interface are reserved for PLC integration.

Signal processing, concentration inversion, and feeding decision-making.

Based on the AT89C52 microcontroller (clock frequency: 12 MHz, Flash: 8 KB) and an industrial control board, it integrates an input layer (AD817 inverting amplifier for processing sensor signals), a processing layer (fuzzy algorithm implemented in C language), and an output layer (74LS245 driving a relay to control the pump). Size: PCB board 100 × 80 mm, total weight: 300g.

Environmental monitoring and safety assurance.

Includes a WIKA A-10 pressure sensor (0–10 bar, installed in the flow path via a three-way connector) and a micro thermistor (temperature feedback, precision: ±0.5◦C). It is also equipped with a buzzer alarm (threshold: 20% LEL) to issue safety alerts when methanol gas concentration exceeds the preset limit. It adopts an overall modular design, supporting hot swapping and compatibility with 10L fermenters (immersion-type porous tube interface). It is fitted with pluggable PCB terminal blocks to enable convenient and stable wiring connections between hardware modules. The display uses an OLED screen + nixie tube (dual-channel for concentration/volume), and the power supply adopts a voltage stabilization module (5V/12V, with a frequency converter supporting PWM speed regulation).

In the fermentation and cultivation of Aureobasidium pullulans, methanol (as a carbon source/inducer) requires stable liquid concentration (optimal range: 0.5–2.0 g/L). However, direct detection of liquid concentration is easily interfered by mycelia and more costly. Therefore, the headspace methanol gas concentration is used to indirectly infer the liquid concentration. The main technical challenges are:

- The impact of high viscosity of the fermentation broth (mycelium concentration: 5–20 g/L) on mass transfer;

- Dynamic changes in liquid concentration caused by the metabolic consumption of Aureobasidium pullulans;

- Fluctuation interference from fermenter temperature (25–30℃), pH (5.0–7.0), and stirring speed (100–500 rpm).

Based on the headspace methanol gas concentration (Cg, unit: mg/m³) detected in the fermenter, calculate the methanol concentration in the liquid (Cl, unit: g/L) in real time, with an error controlled within ±5%.

The model is derived from Henry’s Law and corrected according to the fermentation environment of Aureobasidium pullulans in the fermenter.

(1) Henry’s Law under ideal condition^[1]

$$ C_{l}^{0} = H\, C_{g} $$

Where:

C_l⁰: Methanol concentration in liquid under ideal conditions (g/L);

H: Henry’s constant (L・L/m³), with a reference value H₀ = 2.3×10³ for methanol at 25℃^[2].

(2) Temperature correction: van’t Hoff equation

When the fermentation temperature deviates from 25℃, Henry’s constant is corrected according to the van’t Hoff equation^[3]:

$$ H = H_{0} \, \exp\left[-\frac{\Delta H}{R}\left(\frac{1}{T} - \frac{1}{T_{0}}\right)\right] $$

Where:

ΔH: Methanol dissolution enthalpy (−3800 J/mol^[4], exothermic process);

R: Gas constant (8.314 J/(mol・K));

T: Actual fermentation temperature (K), T0 = 298.15 K (25℃).

(1) Stirring speed correction

Stirring speed directly affects the gas-liquid contact area, which is corrected using an empirical formula:^[5]

$$ f(N) = 0.85 + 0.0005\,N $$

Where:

N: Stirring speed (rpm), range: 100–500;

f(N): Mass transfer efficiency coefficient (0.9 at 100 rpm, 1.1 at 500 rpm).

(2) Fermentation broth viscosity correction

To eliminate the impact of mycelium concentration, correction for fermentation broth viscosity is required^[5]:

$$ f(\mu) = \frac{1}{1+0.02\mu} $$

Where:

μ: Viscosity of fermentation broth (mPa・s), detected by a viscometer. In Aureobasidium pullulans cultivation, it is typically 5–50 mPa・s (viscosity of sterile water is 1 mPa・s)^[6].

(3) Aeration rate correction (affecting gas renewal rate)

The aeration rate affects the calculation of the gas renewal rate. A standard aeration rate Q0 is generally taken as a constant, and this model corrects it to better match the actual cultivation scenario of Aureobasidium pullulans:

$$ f(Q) = 0.9 + 0.05\, \frac{Q}{Q_{0}} $$

Where:

Q: Actual aeration rate (L/min);

Q₀: Standard aeration rate (usually 5 L/min.^[7])

(1) Methanol consumption rate model of Aureobasidium pullulans: Monod equation^[8].

$$ r = r_{\max} \cdot \frac{C_{l}}{K_{s}+C_{l}} \cdot X $$

Where:

r: Methanol consumption rate (g/(L・h));

r_max: Maximum consumption rate (0.3 g/(g・h))^[9], parameter for the logarithmic phase of Aureobasidium pullulans);

K_s: Half-saturation constant (0.2 g/L);

X: Cell concentration (g/L), converted via OD600 detection (OD600 = 1 corresponds to X = 0.5 g/L).

(2) Dynamic concentration correction

To account for the liquid concentration lag caused by consumption, a time differential term is introduced^[10]:

$$ C_{l} = \frac{ H\, C_{g}\, f(N)\, f(\mu)\, f(Q) }{ 1 + \dfrac{ r\, \Delta t }{ C_{1}^{0} } } $$

Where:

Δt: Detection interval (h, usually 0.5–1 h).

Model calibration is mainly achieved through on-site calibration, which includes the following three steps:

1. Synchronous detection: During different stages of the fermenter (lag phase, logarithmic phase, stationary phase), simultaneously detect the headspace C_g and liquid C_l (high-performance liquid chromatography after centrifugation to remove mycelia);

2. Parameter optimization: Adjust the coefficients of f(N), f(μ), and f(Q) using the least squares method to ensure the deviation between the model-calculated value and the measured value is < 5%;

3. Metabolic parameter update: In the logarithmic phase (X > 5 g/L), correct r_max to 0.35 g/(g・h); in the stationary phase (X > 15 g/L), correct it to 0.1 g/(g・h).

5.1 Hardware Achievements

A prototype has been built, capable of realizing basic monitoring and control functions. The GT-CX sensor, data transmission, OLED display, alarm, and other components have been successfully integrated into the circuit board. After soldering with a soldering iron and filter debugging, it supports sampling at 5-second intervals. A pumping interface is reserved, compatible with MP series micro-plunger pumps (flow rate: 12.66–25.32 mL/48 h) and supporting intermittent feeding (i.e., pulse mode).

In testing, the device can accurately detect gas-phase concentrations of 50–200 ppm, with an inferred liquid-phase error of 8% (verified in a 10L simulation tank).

5.2 Software Achievements

A gas-liquid conversion model has been developed and presented in software form, enabling real-time visualization of gas/liquid concentration curves (update interval: 5 seconds, retaining 100 historical data points). Data storage: CSV export, supporting offline analysis. The overall software supports signal filtering (low-pass, cutoff frequency: 1 Hz) and threshold alarm (60 g/L).

A fuzzy algorithm has been added to the microcontroller to realize adaptive feeding, improving the fault tolerance of methanol feeding. Tests show a response time of 10 seconds.

6.1 Problem 1: Unstable Gas-Liquid Conversion Calculation

During dynamic fermentation, the gas-liquid conversion calculation is unstable, with a maximum error of up to 10% and an average error of approximately 6%. The model deviation is larger at high cell concentrations or high methanol concentrations. Specific manifestations:

1. At high cell concentrations (X > 20 g/L), the viscosity exceeds 50 mPa・s, leading to an increased error in the mass transfer correction coefficient f(μ) (±8%).

2. When the methanol concentration exceeds 3 g/L, Aureobasidium pullulans is inhibited, requiring the addition of an inhibition term for correction.

Analysis indicates that the error mainly stems from signal transmission interference, model simulation deviation, and local temperature error (±2◦C).

For hardware, consider optimizing the filter and amplifier in the signal transmission circuit, and adding a crystal oscillator circuit to reduce offset.

For the gas-liquid conversion model:

- Combine an online turbidimeter to monitor X in real time, replacing offline OD detection to shorten the lag time.

- Add integrated NTC thermistor feedback to update the f_T factor in real time.

- Expand the model to include volume dynamic compensation (based on an ultrasonic liquid level gauge, budget: approximately 300 RMB).

- Introduce a Kalman filter to improve robustness.

- Establish a machine learning correction model and train an error compensation algorithm using historical data.

6.2 Problem 2: Sensor Drift and Calibration Requirements During Long-Term Use

After using the catalytic element for more than 6 months, the zero-point estimation drift exceeds 5%, requiring manual calibration.

Embed an automatic calibration routine in the software (weekly standard gas verification), combined with manual adjustment via a remote control (supported by GT-CX). Alternative solution: Replace with FFKM sealing material to extend the service life to 12 months, though this increases costs.

6.3 Problem 3: Insufficient Expandability

Currently, it is only compatible with 10L fermenters; for fermenters of other capacities, the hardware size needs to be redesigned.

Modular upgrade, reserving multi-channel interfaces (Watson–Marlow 323S pump as backup); parameterize the software to support scale adjustment.

1. [1] Henry, W. (1803). Experiments on the quantity of gases absorbed by water, at different temperatures, and under different pressures. Philosophical Transactions of the Royal Society of London, 93, 29–42. https://doi.org/10.1098/rstl.1803.0004
2. [2] Sander, R. (2023). Compilation of Henry's Law Constants (Version 5.0.0) for Water as Solvent. Atmospheric Chemistry and Physics, 23(17), 10901–12440. https://doi.org/10.5194/acp-23-10901-2023
3. [3] Lau, K., et al. (2010). Modeling the temperature dependence of the Henry's law constant of organic solutes in water. Fluid Phase Equilibria, 290(1–2), 166–180. https://doi.org/10.1016/j.fluid.2009.11.020
4. [4] National Institute of Standards and Technology (NIST). Thermodynamic Properties of Organic Compounds Database (2025). Retrieved from https://webbook.nist.gov/chemistry/ (DOI: 10.18434/T4D303)
5. [5] Sieblist, J., Büchs, J., & Takors, R. (2020). Scale-up of stirred-tank bioreactors based on k_La correlation and power input. Biochemical Engineering Journal, 163, 107752. https://doi.org/10.1016/j.bej.2020.107752
6. [6] Luan, X. S., Luan, X. Y., & Zhang, C. K. (2024). Factors influencing fermentation and bioactivity of melanin glucan from Aureobasidium pullulans LHS-m022. Journal of Peking University (Natural Science Edition), 60(5), 799–806. https://doi.org/10.13209/j.0479-8023.2024.044
7. [7] Chen, J., & Du, G. C. (2021). Principles and Technology of Fermentation Engineering. Beijing: Higher Education Press, pp. 189–192.
8. [8] Monod, J. (1958). Recherches sur la croissance des cultures bactériennes. Paris: Hermann et Cie.
9. [9] Yang, Y., Ndabacekure, O., Liu, W. X., Xu, X. R., & Zou, X. (2025). Application of Aureobasidium resources: Biomanufacturing and sustainable development. Acta Microbiologica Sinica, 65(4), 1695–1713.
10. [10] Bird, R. B., Stewart, W. E., & Lightfoot, E. N. (2002). Transport Phenomena (2nd ed.). Wiley.