An Introduction to Online Fermentation Monitoring: From Single-Point Data to Holistic Assessment

Many people new to fermentation have a misconception: the more data points displayed on a fermenter, the greater their control over the entire cultivation process. However, what truly determines the success or failure of the process is not the amount of data recorded by the system, but whether that data accurately reflects the actual state of the microbial culture. All online data initially comes from various sensors installed on the fermenter. These sensors detect physical and chemical parameters within the culture system; this data then passes through transmitters, data acquisition modules, and control systems to ultimately form the curves we see. Understanding these sensors involves more than just grasping their operating principles; it is even more important to know when they are prone to distortion, when they are prone to drift, and when they may mislead process decisions.

Temperature Sensor

Temperature is a parameter that must be monitored in virtually every fermentation system, yet it is also the one that is most easily overlooked among all online data. Many fermentation process engineers believe that as long as the controller consistently displays 36°C, the culture medium must be maintained at 36°C. In reality, this figure represents only the temperature at the sensor’s location and does not necessarily reflect the actual temperature of the entire fermentation system. This is particularly true in scaled-up production, where the interior of a culture system—ranging from tens of liters to hundreds of liters or even several metric tons—is not a completely uniform environment; heat generation, transfer, and dissipation are constantly occurring. Cell metabolism continuously releases heat, the jacket continuously removes heat, and agitation redistributes the heat. When agitation efficiency decreases, culture medium viscosity increases, or local flow patterns change, it is entirely possible for temperature variations to exist at different locations within the same fermenter.

Pt100 platinum resistance thermometers are widely used as temperature sensors in industrial fermentation. The principle behind them is straightforward: they utilize the property of platinum, where its resistance changes with temperature, to convert temperature into a resistance signal, which is then converted into a digital signal by a transmitter. Compared to thermocouples, the Pt100 offers advantages such as good repeatability, high long-term stability, and high measurement accuracy, making it virtually the standard configuration for modern biological fermentation equipment. In fermentation equipment, Pt100 sensors are typically not immersed directly in the culture medium but are instead installed within a temperature-measuring sheath welded to the tank. This design allows the sensor to withstand steam sterilization at 121 °C while facilitating future replacement and maintenance. However, precisely because of this additional sheath, it takes some time for the temperature of the culture medium to be transmitted to the sensor. If the sheath wall is too thick, contains air pockets, or has poor thermal conductivity, the sensor’s response will be slowed. Many people rarely notice this effect in the laboratory because 5-liter fermenters have a small culture volume and rapid flow, allowing the temperature to become uniform almost instantly. However, as equipment scales up, this temperature lag becomes increasingly noticeable.

Another issue often overlooked in the field is the installation location of the temperature probe. In theory, the sensor measures the culture medium temperature; however, if the probe is installed improperly—for example, near the jacket, close to the cooling fluid inlet, or in a stirring dead zone—it is likely to measure only the local temperature rather than the average temperature of the entire culture system. Therefore, even when the displayed temperature is the same, the actual culture medium temperature may not be entirely consistent across equipment from different manufacturers or with different installation methods.

Many resources state that a four-wire Pt100 system can improve measurement accuracy; while this is true, it can easily lead to misunderstandings. The four-wire system primarily addresses errors caused by lead resistance; it does not determine the ultimate accuracy of the entire temperature measurement system. In large-scale fermentation equipment, the wiring distance between the sensor and the control cabinet is often tens or even hundreds of meters. Since the resistance of the leads themselves varies with ambient temperature, a standard two-wire system cannot distinguish between resistance originating from the Pt100 and that from the leads. In contrast, the four-wire system uses two wires to provide a constant current excitation, while the other two wires independently measure the voltage, thereby virtually eliminating the influence of wire resistance. However, the ultimate accuracy a temperature measurement system can achieve depends on the grade of the Pt100 itself, the performance of the transmitter, the accuracy of the data acquisition module, and whether the entire system has been properly calibrated—not merely on whether a four-wire system is used.

Those with practical experience in fermentation typically won’t immediately adjust PID parameters due to a temperature deviation of 0.2 °C. Instead, they first verify whether the probe is loose, the sheath is contaminated, the calibration is correct, and the installation location is appropriate. This is because, in fermentation, accurate measurement is always more important than fast control.

pH Electrode

Composite glass pH electrodes are commonly used for online monitoring in industrial fermentation. Consisting of a glass measuring electrode and a reference electrode, they essentially form an electrochemical measurement system. The glass membrane selectively responds to hydrogen ion activity; when the hydrogen ion concentration changes on either side of the membrane, a potential difference is generated. This potential difference is then amplified by a high-impedance transmitter and ultimately converted into the pH value displayed in the control system.

Textbooks typically introduce the Nernst equation, which states that at 25 °C, a one-unit change in pH corresponds to a potential change of approximately 59 mV. While these theoretical concepts are certainly important, it is even more important to understand why an electrode that is perfectly calibrated becomes increasingly inaccurate once fermentation begins. The answer usually lies in the culture medium itself. Fermentation media are far more complex than laboratory buffers. Proteins, polysaccharides, lipids, cell debris, and various metabolic byproducts continuously adhere to the surfaces of the glass membrane and the liquid junction, hindering ion exchange. As contamination worsens, the electrode’s response speed begins to decline, the zero point gradually drifts, and measurement errors continue to accumulate. At the same time, industrial fermentation equipment requires repeated in-line sterilization at 121°C. High temperatures accelerate the aging of the glass membrane and promote the depletion of the electrolyte inside the reference electrode, causing the reference potential to gradually become unstable. Therefore, while the same electrode may respond sensitively when newly installed, after dozens of sterilization cycles, even if it passes calibration successfully, the actual measurements may already exhibit significant lag.

The following phenomena are commonly observed in the field: after adding ammonia solution, the pH curve shows almost no change, but suddenly rises rapidly a few minutes later; or the control system has stopped adding alkali, yet the pH continues to change. Many people’s first instinct is to suspect the control program, but in most cases, it is actually due to a significant decline in the electrode’s response speed. Therefore, when analyzing pH curves, one should not focus solely on the curve itself but should make judgments in conjunction with other online data. If the pH decreases while carbon dioxide in the exhaust gas continues to rise and dissolved oxygen steadily declines, this often indicates active bacterial metabolism; if only the pH changes while other parameters—such as temperature, dissolved oxygen, and exhaust gas—show virtually no response, the electrode’s condition should be checked first, rather than rushing to modify the process.

Maintenance of pH electrodes is also far more important than many people realize. Standard two-point calibration is merely the most basic task; what is even more critical is long-term monitoring of changes in electrode slope, zero-point drift, and response time. When the slope continues to decline and the response becomes increasingly slow—even if the electrode can still pass calibration temporarily and continues to be used—every reading displayed by the control system may deviate further and further from the actual pH of the culture medium.

Dissolved oxygen electrode

If fermentation engineers were asked to choose just one online parameter to monitor, many would likely select dissolved oxygen (DO). The reason is simple: microbial cells consume oxygen almost constantly, and the oxygen supply capacity directly determines the metabolic rate the cells can achieve. For the vast majority of aerobic fermentations, the dissolved oxygen curve spans nearly the entire cultivation process—from inoculation, through the exponential growth phase, feed addition, and biomass accumulation, all the way to product synthesis—with each stage leaving its mark on this curve. Precisely because of this, many beginners tend to view dissolved oxygen as a “measure of microbial vitality.” When they see DO drop, they assume the bacteria are growing well; when they see DO rise, they think there’s a problem with the bacterial population. In fact, dissolved oxygen is never a direct indicator of bacterial status; rather, it is the result of the combined effects of oxygen supply and oxygen consumption.

Let’s take the most common example. In batch fermentation with feed additions, it’s common to observe the following phenomenon: dissolved oxygen, which had been stable at 30%, suddenly rises rapidly to 80% or even 100%. Many people’s first reaction is that the oxygen supply has increased, but the real reason is often that the feed addition has been interrupted and glucose has been depleted, causing the bacteria to rapidly reduce their oxygen consumption rate due to a lack of carbon source. The oxygen supply hasn’t increased; it’s simply that the bacteria are no longer consuming oxygen at a high rate, so the dissolved oxygen in the culture medium quickly rebounds. Conversely, when feed addition resumes, the bacteria resume metabolism, the oxygen consumption rate increases again, and DO drops rapidly once more. Therefore, changes in DO reflect the balance between supply and demand rather than the oxygen supply capacity itself.

Currently, the most common type of dissolved oxygen electrode used in industrial fermentation is the polarographic dissolved oxygen electrode. Its core structure consists of a platinum cathode, a silver/silver chloride anode, and an oxygen-permeable membrane covering the exterior. During operation, a stable polarization voltage must be applied to the electrode. Oxygen from the culture medium diffuses through the membrane into the interior of the electrode, where a reduction reaction occurs at the cathode, generating a weak diffusion current. The more oxygen present, the greater the diffusion current. After conversion by a transmitter, this is ultimately displayed as the DO value in the control system. There is one issue that is often overlooked here. Many people believe that the DO probe directly measures the oxygen concentration in the culture medium, but this is not entirely accurate. What the electrode actually detects is a current signal related to the partial pressure of oxygen. After calibration, the control system typically displays this as a percentage of air saturation; it can also be converted to mg/L based on conditions such as temperature and pressure. Therefore, even for the same DO reading of 50%, the corresponding actual dissolved oxygen concentration is not exactly the same under different temperature and pressure conditions. This is precisely why DO must be recalibrated before the start of each fermentation.

Typically, prior to inoculation, under constant agitation and aeration conditions, air saturation is used as the 100% calibration point; the zero point is established through nitrogen purging or the sodium sulfite oxygen consumption method. Only after calibration is complete can the control system correctly convert DO changes throughout the fermentation process. However, even after calibration, this does not mean that all data is reliable. The most significant characteristic of DO is its response lag. Oxygen must first diffuse from the culture medium through the oxygen-permeable membrane, then diffuse into the electrode, and finally undergo an electrochemical reaction; therefore, the electrode always requires a certain response time. While the response speed of modern optical fluorescence-based dissolved oxygen electrodes has improved significantly, the T90 response time for traditional polarographic electrodes typically still takes several tens of seconds. When DO in the culture medium changes dramatically, what the control system actually observes is what occurred several tens of seconds earlier.

For 5-liter laboratory-scale fermentation, this delay typically has little impact. However, in high-density fermentation, a few tens of seconds are sufficient for the cells to undergo a noticeable metabolic change; therefore, many process engineers mistakenly assume that PID control is unstable, when in fact it is simply the natural lag inherent in the sensor itself. Additionally, DO has another characteristic: it is easily affected by flow patterns. If the stirring speed decreases or the culture medium’s viscosity increases, the rate of fluid renewal around the electrode slows down, causing the oxygen diffusion boundary layer to thicken. Even if the overall DO of the culture medium remains unchanged, the rate of oxygen diffusion near the electrode will decrease, ultimately resulting in an underestimated measurement. Therefore, in high-viscosity fungal fermentation or high-density cell culture, DO readings are often more sensitive than the actual oxygen conditions in the cells’ environment. Truly experienced process engineers rarely analyze DO in isolation. They prefer to analyze DO in conjunction with feed rate, air flow rate, oxygen concentration in the exhaust gas, and carbon dioxide concentration, because only by combining these data can one truly determine what is actually happening to the cells.

Pressure Sensor

Compared to temperature and dissolved oxygen, tank pressure seems to attract less attention. Many people believe that pressure sensors are only used to trigger overpressure alarms and that as long as the readings remain stable, everything is fine. In reality, however, pressure affects nearly all gas-related online data. The most direct impact is on oxygen. Many process engineers have encountered the following situation: to increase oxygen supply capacity, the tank pressure was raised from 0.03 MPa to 0.05 MPa, resulting in an immediate and significant increase in DO. Some attribute this to a sudden decrease in oxygen consumption by the microbial cells, but in reality, it is largely due to the increased oxygen solubility in the culture medium following the rise in pressure. Therefore, when analyzing DO changes, ignoring pressure variations can easily lead to erroneous conclusions.

Industrial fermentation typically employs diaphragm-type pressure transmitters for measurement. Since the equipment must undergo steam sterilization at 121 °C, and the high-temperature culture medium cannot come into direct contact with the sensor, a diaphragm-sealed design is generally used. The pressure of the culture medium first acts on a flexible diaphragm, then transmits the pressure to the interior of the sensor via a fill fluid, thereby preventing direct damage to the measuring element in high-temperature, high-humidity environments. In many cases, what truly affects the accuracy of pressure measurement is not the sensor itself, but rather the installation method. If the pressure tapping port is located in an area where steam tends to accumulate, the continuous formation of a liquid column from condensate may introduce additional static pressure; if the tapping port is directly opposite the stirrer blades, pulsations generated by high-speed fluid flow may superimpose on the pressure signal, causing the curve to fluctuate constantly. Therefore, high-quality fermentation equipment often avoids areas of strong turbulence as much as possible and minimizes condensate accumulation.

For process engineers, pressure is particularly significant because it is used in the calculation of many other variables. For example, DO calibration, oxygen concentration conversion in exhaust gas, oxygen transfer capacity calculations, and OUR and CER estimates all rely on pressure parameters. Pressure itself may not be the most critical data point, but ignoring it can cause many other data points to deviate from reality.

Rotational Speed Sensor

Agitation speed is typically one of the most frequently adjusted control variables during the fermentation process. While many people view it as a simple mechanical parameter, it actually determines not only how fast the impeller rotates, but also the mixing efficiency, oxygen transfer capacity, heat transfer efficiency, and shear environment of the entire fermentation system. When dissolved oxygen (DO) levels drop, the most common response of the control system is to increase the rotational speed. As the rotation speed increases, the flow of the culture medium intensifies, bubbles are broken into finer particles, the gas-liquid contact area expands, and the oxygen transfer coefficient (kLa) consequently increases, allowing DO to stabilize once again. Therefore, in most automatic control strategies, rotation speed serves as a key means of maintaining DO.

Industrial equipment typically uses Hall effect sensors or magnetoresistive sensors to measure rotational speed. A speed-sensing gear or magnet is mounted on the stirrer shaft; each full rotation generates a fixed number of electrical pulses, and the control system calculates the actual rotational speed based on the number of pulses received per unit of time. This measurement method is highly reliable in itself, but there is another issue that requires attention. The fact that the control system displays 800 rpm does not necessarily mean that the culture medium is actually achieving the mixing effect corresponding to 800 rpm. As microbial cells continue to grow, the viscosity of the culture medium gradually increases; particularly during the cultivation of fungi, actinomycetes, and high-density cell cultures, the fluid properties undergo significant changes. At the same 800 rpm, sufficient turbulence may be generated in the early stages of fermentation, but by the later stages, much of the energy has been dissipated due to viscosity, resulting in a significant decline in mixing efficiency. Therefore, for large-scale fermentation, process engineers are not primarily concerned with the rotational speed itself, but rather whether the power input per unit volume, oxygen transfer capacity, and shear intensity meet the needs of the microbial cells.

Exhaust Gas Analyzer

If DO reflects the balance between oxygen supply and demand, then the exhaust gas analyzer provides a view of the actual respiratory process of the microbial cells. In recent years, an increasing number of fermentation systems have begun to incorporate exhaust gas analyzers—not merely to record oxygen and carbon dioxide concentrations, but because this data reveals information that traditional sensors cannot directly observe.

Microbial growth requires the continuous consumption of oxygen while releasing carbon dioxide. Consequently, the continuous decline in oxygen concentration and the continuous rise in carbon dioxide concentration in the exhaust gas are the most direct indicators of metabolic activity. Compared to DO, the greatest advantage of exhaust gas analysis is that it is virtually unaffected by electrode fouling, culture medium viscosity, and local flow patterns, providing a more accurate representation of the average metabolic state of the entire fermentation system.

Modern exhaust gas analysis typically employs paramagnetic oxygen analyzers to detect oxygen concentration, utilizing the paramagnetic properties of oxygen molecules to perform measurements; Carbon dioxide is most commonly measured using non-dispersive infrared (NDIR) technology, which calculates concentration based on a characteristic absorption peak near 4.26 μm. Both methods offer high stability and are therefore increasingly used in industrial fermentation. However, the true value of exhaust gas analysis lies not in the oxygen and carbon dioxide concentrations themselves, but in the derived parameters: OUR (oxygen uptake rate), CER (carbon dioxide release rate), and RQ (respiratory quotient). These metrics provide a more direct reflection of the microbial metabolic state and have become key data sources of focus in recent years for PAT (Process Analytical Technology) and smart fermentation.

Many process engineers have encountered this scenario: while dissolved oxygen (DO) remains stable, the CO₂ concentration in the exhaust gas begins to decline steadily. In such cases, the actual change is not in oxygen supply, but rather that microbial metabolism has begun to weaken. Continuing with the original feeding strategy can easily lead to carbon source accumulation or even metabolic imbalance. Consequently, in an increasing number of smart fermentation systems, exhaust gas is no longer merely an auxiliary monitoring parameter but is gradually becoming a key basis for assessing the state of the microbial population. Truly skilled process engineers rarely rely on a single curve to make judgments. Instead, they are accustomed to observing DO, exhaust gas, air flow rate, feed rate, and motor speed together on a single trend chart. This is because what truly reflects the fermentation process is never a single sensor, but rather the constantly changing relationships among multiple data points.

Biomass Sensor

For fermentation engineers, the parameter they most want to know in real time is often neither temperature nor pH, but rather how much the microorganisms have grown. Unfortunately, biomass remains one of the most difficult parameters to measure online during the fermentation process. In the laboratory, we can take samples every few hours to measure OD600, or determine wet weight, dry weight, and viable cell count, and then plot growth curves based on incubation time. Although these methods are accurate, they all share a common problem—offline testing means the data is already lagging. By the time the test results are available, the state of the microbial culture may have already changed.

In high-density fermentation, a few hours is enough for the cells to transition from the exponential growth phase to the stationary phase; for processes with strict metabolic control requirements, the feed addition window may be as short as ten to twenty minutes. If we continue to rely on manual sampling, many critical time points are often missed. Therefore, online biomass monitoring has long been one of the most sought-after breakthroughs in the field of fermentation.

Currently, the methods most commonly used in industry are the optical method and the capacitance method. The principle behind the optical method is relatively easy to understand. As the number of cells in the culture medium increases, light passing through the medium is scattered and absorbed, causing a change in the light intensity reaching the detector. By establishing a calibration model, the scattered signal can be converted into the corresponding cell concentration. For microorganisms with relatively uniform cells, such as yeast and E. coli, this method typically yields good results; consequently, many laboratory fermenters are now equipped with online turbidity or online OD monitoring systems. However, once actual production begins, the situation is not so straightforward. The sources of signals in the culture medium are not limited to cell bodies. Bubbles, the color of the culture medium, protein precipitates, flocs, mycelial clumps, and even changes in the color of the culture medium itself can all alter the path of light propagation. This is particularly true for filamentous microorganisms such as fungi and actinomycetes; once their hyphae continuously entangle to form mycelial balls, the light scattering patterns differ completely from those of bacteria. Even with the same OD value, the actual biomass may have changed significantly. Therefore, online OD is, more accurately, a process signal related to biomass rather than cell concentration itself.

Capacitance-based online biomass detection is increasingly being adopted for industrial high-density cultivation. This method does not rely on light but instead utilizes the capacitive properties of living cell membranes. When an alternating electric field is applied to the culture medium, living cells with intact cell membranes become polarized, producing a distinct dielectric response, whereas dead cells, culture medium, and inorganic particles generate virtually no such signal. Therefore, capacitance-based methods actually measure live cells with intact cell membranes, rather than OD or dry weight in the traditional sense. This is why many people find the data to be inconsistent when they first encounter capacitance-based methods. In fact, the two methods measure fundamentally different things: one focuses on light scattering, while the other focuses on live cells.