Description of the fertilizer seeder

The fertilizer seeder used in this research was designed to work under conservation tillage conditions, it is marketed with one or up to eight bodies, each body has a pneumatic seed dosage system, a fertilizer dispenser and solid pesticides. It uses a toothed cutter disc, which functions as a driving wheel, transmission by chains and gears to drive the dosers, in the case of the fertilizer doser the transmission by chains and gears move to a pinion and this in turn to the crown, on which is fixed a bar located inside the hopper (Figure 2). Considering the operating speed of the machines for the application of inputs, with maximum speed of 12 km h, at this speed and with the settings of the machine selected for this investigation, the cutter disk would rotate at a maximum rotation frequency of 146 rev/min, the bar-disk dosing of fertilizer at 35 rev/min, and the grooved roller for dosing pesticide at 144 rev/min, these data were considered as design specifications and to perform the mechatronic system tests.

Figure 1 Fixing the CD motors to the feeders. 

Figure 2 Schematic of the electronic circuit and control logic. 

Mechanical subsystem

The SM was adapted to a body of the fertilizer seeder. To operate the granulated pesticide dispenser, the EMG49 gearmotor, with 24 volts of direct current (VCD), was used, with a rotation speed in vacuum of 160 rev/min, maximum torque of 1.6 Nm and power of 34.7 W. With an encoder integrated 245 pulses per turn of the arrow on each output line A and B. The arrow of the motor is coupled directly to the pesticide dispenser (Figure 1b).

To actuate the dispenser fertilizer one gearmotor 350W was used to 24VDC, with two output arrows, one from the reducer and the other direct from the motor. The arrow with reduction reaches a frequency of rotation in empty of 500 rev/min, and the arrow without reduction of 6 255 rev/min; in the latter, the set encoder was YUMO E6B2-CWZ3E to 1024 pulses per revolution in each output line A and B, which is fed between 5 to 15 VDC. The exit arrow with reduction drives a pinion-crown gear, by means of a chain transmission between the arrow of the pinion and the arrow of the gearmotor, in the crown the bar-disc that gives movement to the fertilizer is fixed (Figure 1a and Figure 2).

Electronic subsystem

The SM is based on the ‘open source’ Arduino development platform for free hardware and software. It uses the Arduino MEGA board (^{Atmel Corporation, 2016a}) as the main control unit, based on the Atmel2560 microcontroller. The following elements were connected to this board: liquid crystal display (LCD 20 x 4), real time clock, SD memory, matrix keyboard and three Arduino UNO boards (^{Atmel Corporation, 2016b}) in I2C communication (Inter-Integrated Circuit), where the Arduino Mega board acts as a master and the Arduino Uno as slaves. To perform the laboratory tests, two more arduino were connected to measure the mass of the test material delivered by the dosers (Figure 2).

An Arduino-Uno was used to control the fertilizer doser and another for the pesticide doser. In each one the following elements were connected: voltage sensor, current sensor, incremental encoder, power unit and the Arduino Mega. Another Arduino Uno was used to measure the forward speed of the planter, to which a YUMO E6B2-CWZ3E encoder and the Arduino Mega were connected. A power unit was designed based on opto-couplers and power transistors, which amplify the PWM control signal sent by the microcontrollers to the motors isolating them from the high voltages. The SM is powered by the tractor's electrical system at a voltage of 24 VDC.

Control subsystem

The process followed for the development of the dosing control system was as follows:

Step 1. Experimentally we identified the equation that relates the pulse width (duty cycle) of the PWM signal to the pulse frequency of the encoder, this separately for each doser. For this, different step inputs of the pulse width were applied (taking the minimum value of 50 and the values to increase by 50 until the maximum value of 1023 was taken) and the pulse frequency obtained from the motor encoder was measured, with each selected pulse width the motor-doser was operated in a time of 10 seconds, taking measurements of the pulse frequency every 98.304 ms. The value of the frequency considered as a response to the applied pulse width of the PWM signal was the average of the measurements obtained from 2 to 10 s. In the case of the fertilizer doser, the test was performed with the motor coupled to the doser with the hopper empty. In the agrochemical doser was carried out under three conditions, the first only with the engine, the second with the engine coupled to the pesticide dispenser with the hopper empty and the third as in the previous case, but with the hopper full of test material and gate completely open.

Step 2. Characterization of the response of the system to a step input. It was obtained as the average behavior in 30 repetitions, of the application of a step input of the PWM signal (with a value of 1023) and its response in pulser frequency of the encoder, in the time interval of the beginning of the application of the input step (zero time) at 1.5 s and making measurements of the pulse frequency at regular intervals of 16.384 ms. The conditions under which the tests were conducted for the fertilizer and agrochemical doser were those described in step 1.

Step 3. Generation of the transfer function of the plant. Using the data obtained in step 2, the transfer function of the plant was determined using the ‘System Identification Tool’ tool of the System Identification Toolbox of the Matlab programming environment, calling this tool from the command window with the ‘ident’ instruction.

Step 4. Tuning the PID controller. With the transfer function of the plant, the tool ‘pid tuner’ of the Control System Toolbox of Matlab was used, making a call to this tool in the command window with the instruction ‘pidtool’, to find the values of the constants kp, ki and kd of the PID controller. The constants of the PID controller were also determined using Equation 1, which represents the transfer function of the PID ( Gc(s)) controller and the transfer function of the estimated plant (Gp(s)), obtaining with them the function Closed-loop transfer of the controller system (Equation 2) ^{Alfaro et al. (2002} y ²⁰⁰³⁾; ^{Anaya et al. (2014)}.

Where:Glc(s) is the closed-loop transfer function, Gps the plant transfer function and Gcs is the controller's function.

In Equation 3 it shows the expression that allows to determine the control signal, by reducing the difference between the reference (frequency of required pulses) and the measured value (measured pulse frequency) of the control variable in the system.

Where: u(t) is the control signal, et is the error between the reference value and the current value kp, ki and kd are the constants of the proportional, integral and derivative controller respectively.

Step 5. Finally, the control algorithms of the mechatronic system were programmed. The control algorithm programmed in the arduino one that controls the fertilizer and pesticide dispenser is the one presented in Figure 3, where ut is determined with Equation 3, for PWMr and PWMm, it is determined with Equation 4 for the fertilizer dispenser and Equation 5 for the pesticide dispenser.

Nomenclature used: n_int = number of interruptions from the beginning of the sowing work until the present moment (dimensionless), G_d = turns given by the seed disk, t_aL = time counted since the beginning of the work (s), ∑i=oi=taLPcRi = total pulses counted in time since the start of the work (pulses), P_cm = pulses counted in the motor encoder (adim.), ∑i=oi=taLPcmi= total pulses counted in the motor from the start of the work (dimensionless), G_d = turns of the dispenser (adim.); F_t = theoretical fertilizer applied during the work (adim.); T_ra = real application time (s), T_ss= time without application (s).

Figure 3 Dosing control algorithm. 

Programming and operation of SM

The SM programming was done in C++ language with the Arduino software version 1.0.5 [22].

The arduino uno (slave one) was programmed to determine the forward speed of the planter (vs), (Figure 2); through, of the counting of pulses generated by the encoder when turning the wheel in an exact time of 163.84 ms, sending the determined value of the speed via I2C communication.

The arduino Mega governs the functioning of the SM. Controlling the mode of operation of the slaves in three modes, which are: inactivity mode, input application mode and test-calibration mode. When the SM is switched on, it starts in idle mode, waiting for the user to change the operating mode; either to the application or test-calibration mode. When one of these modes is selected, it activates the corresponding slaves to execute the selected operating mode and sends the data required by each slave for its operation. When the input application mode is active, the arduino mega expects the signal from the slave one that indicates that the speed of displacement of the planter has been updated, at which time it updates each one of the operation data of the slaves and extracts from them via I2C the data that characterize the performance of the dosers, in addition to checking and updating the operation data entered by the user or changing the operation mode.

When the SM operates in test-calibration mode, the user enters the rotation frequency to which the dispensers must turn and the number of turns they must make. Once active, the mega Arduino controls the rotation frequency of the dosers with the selected magnitude and stops them when they have made the turns chosen by the user, receiving the data of importance of each slave every 163.84 ms via I2C communication.

The arduino uno (slave two) was programmed to control the operation of the fertilizer doser and arduino uno (slave four) of the pesticide dispenser (Figure 2). Its operation starts with the configuration and initialization of each of the devices connected to it and waits for the master to assign the operating mode, either in application, test-calibration or inactivity mode; for each operating mode it receives from the master the data required to execute it and implements the dosing control algorithm (Figure 3). When the application mode is active, it expects the master to update the operation data and send the most important data of the application task, which is done every 163.84 ms. In the case in which the active mode of operation is that of calibration tests, it waits to receive updated operational data from the master and returns the most important data resulting from the activity, this is done until the condition of unemployment is met.

The algorithm implemented in the slaves that control the dosers is presented in Figure 3, which starts with the input data sent by the master, which are: the advance speed of the planter (vs), the pulses generated by the encoder at each rotation of the motor shaft (Pem), the amount of material delivered at each rotation of the dispenser (q), the desired application rate (Q) and distance between rows (dsur), and for when operating in test mode-calibration also sends the number of turns to be given in the doser and the frequency of rotation to which you must perform them. The slaves perform the counting of pulses (Pcm) of the encoder fixed in the arrow of the motor in an exact time of 163.84 ms Tcm) thus determining the frequency of pulses generated by the motor encoder (fm) and of the received data. determines the frequency of rotation required (fr), then with the required pulse frequency the required pulse width is estimated (PWMr) and with the measured pulse frequency the measured pulse width is determined (PWMm) . The pulse width required or measured for the case of the fertilizer doser is determined with Equation 4; and for the pesticide dispenser with Equation 5.

Where: PWM is the required or measured pulse width of the PWM (dimensionless) signal and f the required or measured pulse frequency, both in the motor that drives the fertilizer doser.

Where: PWM is the required or measured pulse width of the PWM (dimensionless) signal and f the required or measured pulse frequency, both in the motor that drives the pesticide dispenser.

The difference between the required pulse width (PWMr) and measured pulse width (PWMm) corresponds to the error value (em) used by Equation 3, which allows obtaining the control signal (ut) of the doser (pulse width of the PWM signal) with which it is fed to the corresponding dosing motor. In addition, the calculation is made of the area in which the input has been applied (Aap) in m², the theoretical applied quantity of the input (F) in grams, the actual working time of the dispenser (Tefs) in percent. The data of importance of the performance of the control of the doser and corresponding to the work done during the work of application of the input or tests are sent to the teacher, data such as effective time of work of the doser (Tefs), area covered during the application (Aap), applied theoretical mass (F), rotation frequency measured in the doser (nm) and control error (en). When the slaves operate in test mode they also measure the voltage and current of the motor supply, and calculate the power and torque (Equation 6) required of the motor.

Where: Vef effective voltage (V); P power to the motor (W); T torque (Nm); w angular speed of the motor shaft (rad s^-1); Dc pulse width of the PWM signal sent (takes values from 0 to 1 023, dimensionless); Vf voltage measured on the line that supplies the motor (V) and I supply current to the motor (A).

Once the programming and the control algorithm were implemented in each of the Arduino, tests were performed to determine the PID control performance using different values of the constants, based on the estimates with the Matlab PID Tuner tool. For each of the constants kp, ki and kd three values were selected, the first group of them used to obtain a response of the control signal that quickly eliminates the error, when the error is greater than 10%, the second group of values is used when the error is in a range of 5 to 10% of the reference value and the third group when the error is less than 5% of the reference value.

Evaluation of the mechatronic system through laboratory tests

The tests were made using a body of a fertilizer-seeder, in the test bench of seeders of the laboratory of the National Center for Standardization of Agricultural Machinery (CENEMA), which belongs to the National Institute of Agricultural and Livestock Forestry Research (INIFAP). The test material used was commercial urea fertilizer (46-0-0), for the fertilizer and pesticide dispenser, based on the Mexican standard NMX-O-222-SCFI-2004 (^{Secretaría de Economía, 2004}) and avoid the handling of toxic products in the case of the pesticide dispenser. The average properties of the test material were as follows: bulk density of 766.8 kg m³, coefficient of static friction of 30°, grain size less than 5 mm and humidity 0.22%.

In the test of each batcher. Five rotation frequencies were used for the fertilizer dispenser (dosing bar) 5, 12, 20, 27 and 35 rev/min; and five for the pesticide dispenser (grooved roller) 10, 32, 55, 77 and 100 rev/min. The opening of the gate that regulates the exit of the material from the hopper was placed in the middle position (50%) and the hopper was placed at a filling level of 80% of its capacity with the test material in both cases. At each rotation frequency four repetitions were made and for each repetition the dosing bar was rotated 20 times and the grooved roller 30 times. In each repetition the frequency of rotation of the dispenser, the control error, the number of turns, voltage and current of feeding of the motor were measured at regular intervals of 114.688 ms, in addition, the amount of test material delivered by each batcher was measured at regular intervals of 645 ms for the agrochemical batcher and 715 ms for the fertilizer.

The voltage and current sensor were verified with a digital multimeter (Model: UD87 from the company URREA Herramientas Profesionales, SA of CV), showing differences in measurements of less than ± 2% in both cases. The load cell was verified with calibrated weights of one, two, three, four and five kilograms obtaining a difference in measurement less than ± 0.05%.