DFC, “self-learning” flow rate control algorithm
Why Flow Rate control performance is so critical?
Many microfluidic applications require monitoring and adjusting the flow rate while an experiment is running. This is often used to control the sample volume dispensed and/or the sample flow rate. When dealing with drug testing experiments, flow-rate control is used for determining the time of exposure or for applying a chemical concentration gradient. In droplet microfluidics, the flow-rate is determines droplet size, frequency, and monodispersity. In cell culture experiments, high flow rates can lead to shear-stress which impacts cell shape and growth.
Combining both pressure and flow-rate control/monitoring provides an extended level of microfluidic flow-control. Indeed, mastering both parameters allows for a chip manufacturer to perform quality testing procedures by calculating hydrodynamic resistance, leakage and burst pressures.
Fluigent systems enable controlling and monitoring both pressure and flow-rate with an easy way of limiting the pressure and/or flow-rate applied to a system. Compared to a syringe pump which continues to perfuse regardless the actual flow rate in the chip channels, flow sensors and Fluigent algorithms have been able to predict the pressure values to apply to reach set flow rates. Working to improve the customer experience, Fluigent now presents its latest enhancement to microfluidic flow rate control: the DFC, self-learning algorithm.
Operating principle: Both pressures and flow-rates are measured by Fluigent pressure controller and the Flow-Rate Platform, the DFC is able to automatically adjust pressure(s) to reach the flow-rate set-point(s).
From traditional to a “self-learning” flow rate control algorithm
The previous paragraph discussed the importance of monitoring and controlling the flow-rate. Here we want to share how Fluigent has evolved from a classic PID to a “self-learning” algorithm to overcome typical issues in microfluidic flow rate control such as calibration, multi-channel interactions and resistance changes during experiments.
Automatic Microfluidic Pressures Control based on flow-rate with a PID controller
While syringe pumps produce perfusion flow-rates from a piston displacement, that can result inflow rate instability, a pressure pump is generally associated with one or several flow sensors. Traditional pressure-based microfluidic flow rate control uses a proportional–integral–derivative controller (PID controller or three-term controller), as a control loop feedback mechanism.
In practical terms a PID controller continuously calculates an error value as the difference between the desired flow-rate set point and the actual measured flow-rate. The algorithm then adjusts the pressure to reduce this error and get closer to the flow-rate set point. A simple advantage of using pressure-based flow-rate control in microfluidics would be when a channel is slightly clogged. Without the PID, control the pressure would remain the same, resulting in a decrease in the flow rate.
The first Fluigent software (2010) had an integrated PID algorithm to restore the measured flow rate to the desired flow rate with minimal delay and overshoot, by altering the pressure applied to the setup. With a syringe pump, the piston would continue pushing with the same strength leading to large increases of pressure inside the chip, sometimes up to the burst pressure.
Response of PV to step change of SP vs time, for three values of Kp (Ki and Kd held constant)
PID’s can be efficient, but need a parameter adjustment depending on the downstream device resistance. Basically, the ideal PID parameters will vary depending on the hydrodynamic resistance of the system (i. e. the microfluidic channel dimensions or the viscosity of the liquid). With this type of algorithm, users have manually adjust the parameters of the equation in order to achieve the desired flow-rate response time and stability. This brings limitations in use when changing between multi-channel setups, or differing resistance over time (e.g. cell culture growth).
Microfluidic Pressure Control based on flow-rate with an Autotuned PID controller (2011)
The first improvement of Fluigent’s flow rate control scheme was focused on the adjustment of the control parameters (proportional band/gain, integral gain/reset, derivative gain/rate) to the optimum values for the desired control response.
Before performing an experiment, the user had to launch an automated calibration procedure where the software would make pressure steps and evaluate at the flow-rate behaviour to automatically calculate optimized PID parameters. With user in mind, Fluigent simplied the use of algorithm with an auto PID tuning and loop optimization software to ensure consistent results.
Fluigent gathered data and develop microfluidic models to suggest optimal tuning based on the setup calibration. Despite this, without some manual adjustments, this method had some limitations when used with multi-channel or resistance changes in setups.
Introduction of the FRCM algorithm (2012)
For a given applied pressure, the value of the flow-rate depends on the microfluidic design of the system. So, as microfluidic designs become more and more complex, the calculation of the required pressure leading to the desired flow rate becomes more and more difficult. The case of coupled, multi-channel chip designs for instance can be challenging to address efficiently.
Introduced in 2012, the Flow Rate Control Module (FRCM) was the first microfluidic algorithm to enable pressure-based flow-rate control on multi-channel microfluidic setups. In simple words, one could say the FRCM is a matrix of PIDs allowing for modeling complex microfluidic channel networks.
Compared to the autotune PID, the FRCM successfully overcomes channel cross-talk situations by predicting and automatically adjusting pressure(s) to reach the flow rate set-point(s). The FRCM was based on a powerful algorithm that enabled the internal modelling of any microfluidic system linking each flow-rate to a combination of pressure orders, whatever the microfluidic design. Thanks to this unique algorithm, the FRCM can predict the pressure values to apply to reach the flow rate set-points even in complex microfluidic systems, such as droplet generators, double encapsulation systems or mass parallel systems.
New self-learning algorithm: DFC (2017)
Though the FRCM overcome many issues encountered in complex microfluidic configurations, it still required calibration and couldn’t perform in a system where hydrodynamic resistance evolves along the experiment. These issues are no longer present with Fluigent’s latest flow control technology update: the DFC self-learning algorithm.
Fluigent Direct Flow-rate Control (DFC) is a self-learning algorithm compatible with all Fluigent pressure controllers. This improvement to the autotuned PID and the FRCM includes a continuous adjustment of the algorithm parameters based on the actual response time and stability Its improved reactivity allows it to counter, in real-time, the interactions between microfluidic channels in complex situations. There are many advantages to direct control:
- Adapted to cell culture or other experiments involving resistance variations: the algorithm adjusts its model to the setup resistance in real-time
- Save precious sample or reagent: reduced time to reach desired flow rates uses less liquid during a calibration step
- Save time with reduced settling time and no calibration
For compatible pressure controllers go to: Microfluidic Components
Flow EZ, flow rate control response to 5µL/min increasing steps each 30 seconds
Impacts of algorithm performance
Algorithm tuning is a difficult and time-consuming process to reach the optimum values for the desired response. For microfluidics and flow rate control stability (no unbounded oscillation) and response time are the two basic requirements. Fluigent new DFC algorithm aims to achieve the best results in both cases.
When used with flow rate sensors, microfluidic pressure controllers are able to maintain a stable flow rate due to lack of moving mechanical parts in this technology. This enables one to achieve the same functionality as a syringe pump, but with a much more stable and reliable output. Increasing the stability of the flow rate will improve the reproducibility of microfluidic experiments. This major factor is essential in many biological and chemical applications for reliable results. Flow rate stability is also essential to control cell shear stress, the tangential component of frictional forces generated at a surface by the flow of a viscous fluid, as it affects the physiology of the cells1. Depending on the field of application, shear stress related flow requirements can be different. (read more on Microfluidic Stability)
In the case of microfluidics, responsiveness is defined as the time it takes for the pressure or the flow-rate in the fluidic reservoir to reach a given set point. Pressure based Microfluidic Flow Controllers display major benefits in terms of responsiveness compared to other types of fluid handling solutions such as syringe or peristaltic pumps. This impacts the experimental time, reducing the elapsed time between the command and the desired flow rate. It also prevents the waste of precious sample.
Any microfluidic pumping system has advantages and drawbacks. Here we present how Fluigent overcame one major drawbacks of pressure pumps: the lack of having optimized direct control of flow-rates. Fluigent has continuously improved its microfluidic technology to develop products with the best performance. Direct Flow Control is an achievement in the field as it combines the advantages of both flow-rate control and pressure actuation: In just one click, a Fluigent pump can switch from flow-rate to pressure control, with the best available flow stability and fast response time without the need for additional settings in very complex microfluidic channel configurations.