1. Introduction
Bioprinting is a booming technology that could revolutionize regenerative medicine [
1]. It has a specific application in medicine of traditional additive manufacturing technology that superposes layers of material to build a biological structure. Although there are several bioprinting techniques, all of them present some common challenges to be solved, such as cell death during and after bioprinting, a long bioprinting time or insufficient micro-vascularization, among others [
2]. Bioinks are commonly cell-laden hydrogels due to their good biocompatibility, so they present highly hydrated 3D networks, such as extracellular matrix, that promote oxygen and nutrient interchange [
3].
It is important to know that hydrogels are very sensitive to external changes (e.g., temperature or humidity) because of their high-water content (80–90%
w/
v). For this reason, several processes can be associated to uncontrolled bioprinting conditions. Some of them are related to the stability and integrity of 3D bioprinting structures [
4] and other ones are focused on the survival of cells during and after the bioprinting process [
5]. In the first block, drying hydrogels will increase the concentration of macromolecules and will promote their crowding [
3]. Furthermore, the rheological properties of hydrogels are affected by temperature, among other parameters [
6]. These two problems are usually controlled during the printing process inside of the bioink cartridge, but unfortunately, the post-printing stage can suffer similar atmospheric-related problems that are currently uncontrolled inside of the current bioprinter enclosure. Hence, the proposed atmospheric enclosure system for bioprinters expects to fill this gap, increasing post-printing stability and, consequently, allowing a higher resolution of the 3D bioprinted structures [
7]. In the second block, cell viability can be affected during both the printing and post-printing processes. In the same way, good cell conditions during the bioprinting process are controlled by bioink cartridges, which are out of the scope of this work [
8]. However, the control of the atmospheric enclosure inside of bioprinters to assure proper conditions in the post-printing stage, promoting high cell viability, is the other main interest of this work [
9].
As far as the authors know, no previous studies have focused on designing, developing and testing an atmospheric enclosure system for bioprinting that could control these critical parameters to assure the integrity and stability of the 3D structure together with the viability of its cells during the post-printing stage. Instead of generating an atmospheric enclosure system, other authors propose to control these atmospheric parameters by reducing bioprinting time, bioprinting into a water-based bath or reducing air flow [
3]. Commercial bioprinters, such as Cellink BioX
® (Cellink; Boston, MA, USA), Poietis
® (Poietis; 33600 Pessac, France) or 3D-Discovery BioFactory
® (REGENHU; 1690 Villaz-Saint-Pierre, Switzerland), control the air flow to minimize biological contamination using High Efficiency Particle Arresting (HEPA) filters [
10,
11,
12]. However, none of them currently solve these issues and the 3D bioprinting process is still relatively slow [
7], so large 3D structures can greatly benefit from our proposed post-printing control of atmospheric conditions, which could turn bioprinters into a temporal bioincubator while the bioprinting process is ongoing [
7,
13].
Proportional integral derivative (PID) controllers are commonly used in atmospheric enclosures of other fields, e.g., neonatal [
14,
15,
16] or egg/bird [
17,
18] incubators. Each different application requires complex or simple adaptations of their mathematical models, according to the factors involved in each of these environments. On the one hand, large enclosures, such as greenhouses [
19,
20], animal buildings [
21] or large spaces [
22], must consider many factors: air convection, wall heat loses, inner heat, humidity generation or solar radiation. On the other hand, small enclosures, such as neonatal [
14,
16,
23] or egg/bird [
15,
24] incubators, can considerer negligible some of these previous factors, but they present a high cross-coupling effect among their variables [
25]. For this reason, some studies have focused their interest only on temperature [
15,
16,
23] or just merge two variables, such as temperature and humidity [
14,
24,
26]. Temperature and humidity are highly related variables, so their simultaneous control is complex and requires a fine tuning of the PID parameters [
27]. Additionally, both temperature and humidity are significantly influenced by CO
2, which increases the complexity of PID controllers due to the cross-coupling effects among these three variables [
28,
29]. It has been observed that the structural stability of Type I collagen cannot be ensured after the bioprinting process without a controlled 37 °C environment [
7]. With this in mind, the CO
2 variable was left out of the scope of this work for two main reasons. Firstly, it is not expected that cells produce significant CO
2 emissions during post-printing, so its concentration should remain stable in the global system. Additionally, the cross-coupling interaction of CO
2 with temperature and humidity will make difficult the optimization of the PID parameters and, consequently, will decrease the robustness and stability of the final PID controller. Hence, we focus our interest on developing a stable and robust PID controller of the two main variables involved in the post-printing stage of bioprinting—temperature and humidity.
Therefore, the main objective of this work is the design, development and validation of a temperature and humidity PID controller for an atmospheric enclosure system with bioprinting purpose. This atmospheric enclosure system could control the post-printing stage independently of the bioprinting technique used. For this purpose, theoretical, simulated and experimental studies have been performed to assure a proper functioning of the proposed system.
2. Materials and Methods
2.1. Atmospheric Enclosure System Design
The designed atmospheric enclosure system is a sub-divided parallelepipedal with different areas: a bioprinting sub-chamber, a climatic conditions generation sub-chamber and an electronic/mechanic components sub-chamber. All designing was done using Autodesk Inventor
®, and different area dimensions are shown in
Table 1.
Figure 1 shows the atmospheric enclosure system prototype with its atmospheric enclosure made of methacrylate. Different methacrylate widths were used in each one of the areas to control heat insulation and minimize conduction heat losses.
Figure 2 shows the different sub-chambers that make up the atmospheric enclosure system.
Bioprinter control is divided into three different processes: bioprinter mechanical control, atmospheric enclosure control and data visualization. All three processes were individually controlled by Arduino® (Arduino s.r.l.; 20900 Monza, Italy) UNO boards, all connected to a Raspberry® (Rapsberry Pi Foundation; Cambridge CB2 1NF, UK) Pi 3, and the whole system was controlled by a personalized Python® version 3.9 (Python Software Foundation; Wilmington, DE, USA) script. This script provides the user full live control over the bioprinter atmospheric enclosure parameters while temperature, humidity and CO2 data are shown.
The control of proper atmospheric conditions inside the enclosure needs generation and sensorization of each one of the inner parameters. In this sense, temperature is generated with two 200-W electrical finned resistances, humidity is produced with cold vaporization of water in a tank using a piezoelectrical transducer and, finally, CO
2 is injected from an external AquaMedic
® (AB Aqua Medic GmbH; Bissendorf, Germany) CO
2 pressurized bottle when a Blau
® (Barcelona Marine Farm S.L.; Barcelona, Spain) (3VA, 14mA, IP 85) electro valve is opened. Likewise, the type and location of every one of the sensors used is shown in
Table 2. Generation electronics is switched ON/OFF by a TONGLING (Xiamen Hongfa Electroacoustic Co; 361021 Xiamen, China) JQC-3FF-S-Z relay moduli.
The main area of the atmospheric enclosure system is the climatic conditions generation sub-chamber. All generation apparatuses were placed in this area to produce the appropriate atmospheric conditions. This sub-chamber is a separated area of the enclosure connected by 4 air inlets/outlets, 2 inlets with axial fans and 2 outlets, as shown in
Figure 3. The working principle of this area is to force the air to enter into the generation sub-chamber where its temperature, humidity and CO
2 are modified and then exit to the enclosure atmosphere again. This process is continuously repeated, creating an air conditioning flow until the atmospheric parameters’ target values are reached.
To ensure a sterile environment at the beginning of the bioprinting process, the bioprinter sub-chamber is sterilized by action of four UV LED panels (420 nm and 440 mW) incorporated in the bioprinter sub-chamber. After this process, air is introduced into the system from outside, passed through a HEPA filter and generates positive pressure to prevent the entry of contaminating external agents.
Although the proposed atmospheric enclosure system has a CO2 pressurized bottle and two CO2 sensors for future purposes, they have not been used in this study.
2.2. Mathematical Modelling
The purpose of this work is to control the inner temperature and humidity of an atmospheric enclosure system. In this sense, the inner heat balance and water mass balance equations are the general non-lineal differential equations used in temperature and humidity control processes [
21,
22]. In this work, terms with a negligible interaction in the total balance were not considered. Therefore, our heat and water mass balance equations [
20] were given by:
where
Cv is the volumetric air heat capacity (J/kg∙°C);
V is the inner air volume (m
3);
Th is the inner temperature (°C);
Ga is the air mass flow that a fan extracts from the chamber to the generation sub-chamber (m
3/s);
is the air density (1.25 Kg/m
3);
Cp is the air-specific heat (1006 J/kg∙°C);
Tc is the generation sub-chamber air temperature (°C);
Qn is the heat dissipation capacity (J/s);
T0 is the external temperature (°C);
R is the enclosure thermal resistance (°C∙s/J);
dn is the inner humidity (g/kg);
dc is the external humidity(g/kg);
Dn is the enclosure humidity gain (g/s).
Transfer functions for the PID controller were obtained applying the Laplace transform to Equations (1) and (2). Defining
X as
these functions are:
where
G1(
s) and
G2(
s) are temperature and humidity terms, respectively.
2.3. Experimental Evaluation
A comparative between theoretical and experimental behaviors (theoretical and experimental transfer functions) of the atmospheric enclosure was proposed. For this purpose, the different experimental behaviors of each parameter (inner temperature and humidity) were tested and determined. To obtain data for the experimental transfer function, single input, single output (SISO) and multiple inputs, multiple outputs (MIMO) tests were performed as follows:
Initial inputs and target values for all tests are shown in
Table 3. Atmospheric values have been chosen based on criteria consulted in the literature [
30].
Experimental transfer functions were obtained from test data using the Matlab® R2018b System Identification Toolbox™. This tool is commonly used to obtain mathematical models of dynamic systems from experimental input and output data. As such, Matlab® R2018b Simulink™ was used to simulate and compare both theoretical and experimental transfer functions.
2.4. PID Controller
Variations in temperature and humidity can change bioprinting materials’ properties and affect the cellular viability, stability and integrity of 3D bioprinted structures. To minimize variations, a PID controller was added to the atmospheric enclosure system. PID controller algorithms are widely used in feedback control systems and are mathematically expressed as [
31]:
A good PID control system can provide the atmospheric enclosure with the necessary stability in the shortest time. In this sense, the Ziegler–Nichols closed-loop method was used in Matlab
® R2018b (MathWorks; Natick, Massachusetts, United States) to tune all PID parameters using process input/output signals. All these calculations were done through a simulation block diagram implemented in Matlab
® R2018b Simulink™, which follows the standard structure shown in
Figure 4.
2.5. Procedure
The workflow was performed as Yinping et al. [
28] in their work. First, mathematical model behavior equations were solved, obtaining all constants needed to calculate the theoretical system transfer functions. Second, temperature and humidity transfer functions were calculated and evaluated with experimental transfer functions obtained from MIMO and SISO tests. Several behavior approaches of the system were obtained using the least-square method in the Matlab
® R2018b System Identification Toolbox™. All possible types in this toolbox were used: one pole (P1), two poles (P2), three poles (P3), one pole and one zero (P1Z), two poles and one zero (P2Z), three poles and one zero (P3Z), two poles sub-damped (P2U), three poles sub-damped (P3U), two poles and one zero sub-damped (P2ZU), three poles and one under-damped zero (P3ZU), one pole and one integrator (P1I), two poles and one integrator (P2I), three poles and one integrator (P3I), one pole and one delay (P1D) and two poles and one delay (P2D) transfer functions.
Next, both transfer functions’ (theoretical and experimental) block diagrams were designed and evaluated. Finally, theoretical and experimental transfer functions were analyzed, obtaining the first Kp (proportional controller), Ki (integral controller) and Kd (derivative controller) values using the Ziegler–Nichols method. Then, this first PID tuning was tested and manually re-tuned to limit overshoots produced by the Ziegler–Nichols method.
To validate the atmospheric enclosure system, several tests were performed using the former calculated Kp, Ki and Kd values. Finally, to check the quality of calculated PID, an experimental error (6) was calculated.
4. Discussion
The objective of this work is to determine the optimum PID controller values of an atmospheric enclosure system for bioprinting. Specifically, we propose to control temperature and humidity to improve the integrity, stability and cell viability of 3D bioprinted structures. In order to design and develop our atmospheric enclosure, three components of the system were analyzed and compared with previous studies: the mathematical model, the transfer functions and the block diagram, but it was also analyzed whether to use coupled or uncoupled variables.
The similarity between the proposed atmospheric enclosure system and greenhouses or animal buildings suggests that the mathematical model used by Daskalov et al. [
20] and Albright et al. [
25] (heat and humidity balances) can be used in our system after some important adaptations. In this sense, the presence of animals in the study of Daskalov et al. [
20] requires some factors in the mathematical model, such as sensitivity to heat or water steam produced by animals and the heat lost due to mechanical ventilation. All of these factors must be eliminated for our bioprinting environment. In addition, Albright et al. included terms of solar radiation heat, ventilation volumetric flow and plant evapotranspiration, which have no purpose in our mathematical model. Lastly, our transfer function calculations from differential equations were based on Yinping et al.’s [
28] work, using a similar structure for transfer functions relating to the behavior of temperature and humidity.
According to the bibliography, there is no agreement about the best way to control temperature and humidity. So, some authors propose coupled controllers [
28,
32,
33] while others propose decoupled controllers [
24,
25,
34]. In this sense, knowing that an increase of 1 °C can decrease humidity by about 2–4 % [
24], we performed SISO and MIMO tests to study each parameter independently and simultaneously, respectively. Considering our design and the MIMO test results (
Figure 8), we finally decided to design a decoupled temperature and humidity controller that produces better results with simple PID calculations as well as increasing stability and robustness of the goal system [
24].
Once the type of controller was chosen, the next step was the composition of block diagram, whose main element will be the system’s plant transfer function for temperature and humidity. Previous studies proposed different types of transfer function to control temperature and humidity: some of them without delay [
14,
32] and others considering delay [
24,
34,
35], which is a parameter totally dependent on the technical specifications (response time) of the electronic components [
21]. Those authors that proposed no delay transfer functions try to simplify the system interactions, but their models are supposed to be less real [
17,
25,
34]. On the other hand, the delay is commonly considered when complex transfer functions with high oscillatory responses and difficulties in stabilizing disturbances are present [
25]. Hence, our system provides a better response using delay, specifically P1D type, similar to other previous studies [
24,
34,
35].
Regarding the block diagram, three different blocks have been used: control, plant and delay. Although this diagram is quite common [
25], in the same way as transfer function, delay inclusion or exclusion will modify the performance of the PID controller. So, some studies excluded the delay for temperature [
17,
22,
28] while others considered it in their schemes for temperature and humidity [
32,
36]. In our case, an adapted version of Yiping et al. [
28] that considers delays has been used, but other complex approaches with predictive algorithms should be considered for future improvements [
36].
Finally, the global performance of our atmospheric enclosure system is pretty good. In this sense, low stabilization time and errors were obtained in the experimental tests. Other authors presented a temperature stabilization time of 2760 s for a 4 °C increment [
14], 600 s for a 10 °C increment [
34] or 246 s for a 5 °C increment [
33]. Nevertheless, our enclosure system can stabilize a 12 °C temperature increment in 311 s. Regarding humidity, previous stabilization times were 420 s for a 10% increment [
33] or 300 s for a 5% increment [
34]. However, our system can stabilize a 40% humidity increment in only 65 s. It is important to note that different settings, such as heat devices, humidifiers or enclosure volume, can unfairly bias this comparison. For this reason, a specific comparative study should be performed when other atmospheric enclosure systems will be available for bioprinting. Meanwhile, we have exposed some differences among our work and the previously exposed studies for better understanding of our previous comparisons. In some cases, no technical specifications of sensors/devices are available; in this case, we cannot go into detail about the reasons of differences [
34]. Other studies with a larger volume of controllers and heaters/sensors are differentially dimensioned [
33]. Maybe the most similar enclosure was an incubator with comparable electronics, but unfortunately, the authors did not control humidity in the analysis [
15].
5. Conclusions
In this work, a PID controller for temperature and humidity control of an atmospheric enclosure system for bioprinting was designed, developed and tested. Theoretical and experimental transfer functions for temperature and humidity were calculated and verified. The results show that the proposed atmospheric enclosure system is capable of stabilizing temperature and humidity in 311 s and maintaining target values with an average error of 1.89% and 1.29% for temperature and humidity, respectively. Hence, the proposed atmospheric enclosure system for bioprinting could improve post-printing environmental conditions to increase the integrity, stability and cell viability of 3D bioprinted structures.
As commented, to guarantee a proper atmospheric enclosure system for bioprinting, a CO2 control is needed in addition to temperature and humidity. In this sense, the next step of this study will be analyzing the system’s behavior after the inclusion of carbon dioxide as well as the interrelation with temperature and humidity. Additionally, bioprinters usually have a heat/cool cartridge in the extruder to control the bioink temperature. So, the addition of this heat/cool source inside the atmospheric enclosure to simulate bioprinter performance while extruding will be another future development.