An Intelligent Approach for Adaptive Liquid Level Control System Using Machine Learning

Abstract

Liquid level control is a mechanism that monitors, compares, and regulates the level of liquids within a process to a set value. Level measurement determines the position of the level relative to the top or bottom of the process fluid storage tank or silo. Level control and measurement are essential to assuring the safety and profitability of industrial processes. In this project, we are Implementing an intelligent mechanism using machine learning to predict the level of liquid in a reservoir based on the time and motor speed thus avoiding errors and keeping the process is ongoing.

Introduction

I. INTRODUCTION

A liquid Level Control System is a system specifically designed to control the level of fluid in tanks. The main aim possessed by these systems is to control the rate at which the pump delivers fluid to the tank so it can reach the desired level inside the tank.

The purpose of the liquid level system is to maintain a specific level of fluid inside the tank. The liquid level control systems find major applications in industrial processes. The controller present in the system generates the control signal that is converted into the desired signal by the actuator and is fed to the plant to perform the desired action.

A. Errors Occur in the Existing Plant

1. Case 1: Flow monitoring
2. Case 2: Orientation of the nozzle- if the nozzle is m misaligned, there would be a loss.
3. Case 3: If there is a blockage in the pipe or nozzle it is very difficult to find the area of blockage or sometimes it is unknown whether the tank is empty or there is a blockage in the pipe due to which the flow of fluid is stopped.
4. Case 4: When the tank gets empty the containers are left unfilled with the liquid and if both the tanks are full the liquid overflows out of the tank.
5. Case 5: When the motor has to be running at a speed of 20 rpm but rotates at 30 rpm then the outflow is more and the tank level will increase.
6. Case 6: Based on the product demand, the quantity of liquid level in the tank has to be ensured.

II. LITERATURE SURVEY

The water level in the tank[1] is controlled according to the set point given by the user,this approach was chosen mainly because fuzzy logic is more accurate and stable for a wide range of set points. The water levels were reached accurately and the time taken to reach the set point depends proportionately upon the value of the error. But the Level in the tank was checked manually. Error Measured Was 2.5. Efficiency was for Different Set Points and Achieved 95.3% 8 Level Set Points Were Used.

Dongsheng Wanga,*, Yan Wang [2] proposes a waterworks optimization control system based on machine learning, through the historical data of the plant water treatment process. offered a new way for optimal control of water, and at the same time confirmed the effectiveness of machine learning in the application of water energy consumption prediction and configuration optimization direction. Second, this paper designs an optimization control system for water plant optimization, consisting of a concise and convenient control interface for the above subsystems, and combines the theory and the use of the water plant optimization control system.This author does not mention the auto correcting of the speed. The efficiency achieved was only 60.9 % .

  Kalidasan, J. Ben Ajai Raja proposed an Automatic Bottle Filling Machine [3] where the Motor Is Used to Rotate the Bottle base plate, and  Sensors were Used to identify the Position of the Base Plate Rotation. Then the Solenoid Valves are used To open the Valve Of The liquid flow to the bottle for a Particular Time, in this work the quantity of flow of liquid in the tank is not considered, it does not have a feedback loop. And No monitoring of level in the tank  and  no automatic liquid filling is considered

Filling Machine which can fill different sizes of containers based on the height[4]. The filling of liquid is done based on the threshold set in the controller. In this work  Identifying the blockages and errors in flow pipes are not considered. and there is no monitoring of the liquid level in the tank.  And no mechanism for auto refilling of the tank was  deduced.

Md. Liton Ahmedl  proposed a n automated bottle filling system using a PLC     based controller [5]. The filling and the movement of the bottle along the plant is controlled by the afore mentioned PLC. The PLC controls a conveyor on which the place unfilled bottle move. The position of the bottles are detected by an infrared sensor and a dc pump is used to  control the flow of water, hence enabling functionality at night. Here  the proposed method doesnot take into consideration the quantity of liquid filled or the monitoring of the level of liquid in the tank. Also  it lacks the implementation of a feedback loop and  a prediction model to estimate  the speed of the motor.

The idea and development of an intelligent tank with a water level sensor [7] is conceptualized wherein the water proportion can be determined by the emptiness and fullness of the tank using artificial intelligence. The proposed system claims the reduction in water wastage but lacks proposing an auto correcting closed loop system. The system also lacks real world efficiency.

The proposed system of ultrasonic transducer based water level sensor using pump switching [8] addresses the problems of untimely response and frequent breakdown of contact sensors due to corrosion and coating from the water medium. Yet it fails to address scenarios where the pump would run unnecessarily when there is no water at the source and overheating of the pump.

Liquid level monitoring using FBG sensor array [9] was also proposed 9 FBG(Fiber Bragg Grating) temperature sensors in a tank of height 100 cm and width 30 cm. The work proposes a level estimation in two steps where level detection is done using classification models and level estimation is done using regression models. The level detection is carried out by calculating under which FBG the liquid level is. The proposed method has error of 3.56 cm to 6.28 cm depending on the machine learning model. It only has a claimed efficiency of 89.50%.

As per our proposed solution, the control of a dc pump using machine learning is a effective and feasible solution. Considering the study done on the comparison between PID control and Neural network control over a dc motor [10], the concept of using newer trained models and logics over conventional PIDs is proven better. Similar work has been done to develop a mechatronic model that monitors the water levels and controls the inflow rate using fuzzy logic [11]. The simulation showed an average 0.345 error with an efficiency on only 75.50%.

The effectiveness of the afore mentioned approach is further emphasized the comparison with fuzzy logics[12], where in the intrustrial controller like PID had an estimated steady state error of 3.56 seconds.

The implantation of IOT for Industrial Monitoring and Control [13] was also proposed. The proposed method relied on the use of a reliable TCP/IP protocol in conjunction with GPRS enabled communication. This was done using a GSM and DTMF module. The proposed system was based on a conventional monitoring system with no auto control or implementation of any prediction models to increase efficiency and effectiveness.

III. OBJECTIVE

The  proposed work is to develop a smart intelligent adaptive system which auto controls the level of liquid in the tank based on the data learn by  prediction models to increase efficiency and effectiveness. The proposed solution is as follows.

IV. PROPOSED SOLUTION

As we can observe from the above-stated literature survey some plants require human interference in the work cell to avoid errors like checking tank level, blockages, flow, leaks, auto refilling, etc. and the systems are not intelligent enough to reduce the error and get the maximum efficiency.and also, these existing system does not consist of any monitoring display to show the parameter values. To overcome these disadvantages, we introduce a hybrid Machine Learning system that can monitor and detects an error in the process and auto-correct it intelligently.

The process is integrated with Machine Learning algorithm[15] which was implemented to control the motor speed for better accuracy and less error,which keeps a record of old input commands and compares them with the given command with the help of a database and suggests the best action to be carried out during the period. Even in case, the command is not given it uses the existing commands from the machine-learned database and executes to solve the particular error.

 This system consists of two modes, semi-automatic mode and an automatic mode wherein semi-automatic mode the human interference is involved and the input is given to the system through speech commands[14] or GUI. in automatic mode, the system analyses the problem and compares it with the database to solves the situation accordingly.

V. CIRCUIT DIAGRAM AND FLOW CHART

1. The Machine Learning algorithm [15]compares with the previous data set at that particular time and corrects the rpm according to it and reduces the error.
2. A system in which the Machine Learning algorithm[15] controls the plant with comparison of the previous database where the algorithms is written in such a way to reduce the error and maximize the efficiency of the production. Where the Machine learning code consist of speech recognition where it filters the unregistered speech frequency through database. The micro-controller displays all the values of the sensors in the display attached to it. The Machine learning system monitors the flow, liquid output and level of the tank.
3. Where to overcome this scenario data sets which consists of day, time, level of water and motors speed are being recorded.
4. Then the recorded data is processed by the ML algorithm[15] and gives the prediction of fluid required to filled the next day.
5. The ML code compares the data of real time output of the motor speed and the reference speed set from the data set by the trained model to detect the error.
6. The ML code adjusts the motor speed to minimize the error according to the reference speed from the data set.

VI. PLANT DESCRIPTION & DIAGRAM

let's assume a  plant where the tanks are filled with liquid and they flow through the pipe goes to the nozzle where the containers come through the conveyor belt and the nozzle valve opens and fills the containers with the liquid, Where to be precise the primary tank is filled with the liquid with the help of a pump and when tank A is empty or goes below the threshold the liquid is pumped to tank A from the secondary tank.

1. The input command is given through voice or manually[15] is executed by the controller.
2. Where the controller is connected to the database and checks or compares the input commands parameters with the previous old commands parameters which are stored in the database.
3. Cu(control unit) will decode the tag, and take appropriate decisions on pump automation.
4. The control unit is connected with different types of sensors and actuators to control the error of the operation and increase the efficiency of the production through uninterrupted services.
5. There are three sensors connected to the tank. L3 ( on top) L2( in middle)  , L1( bottom).
6. As shown in Fig 2, If L1, L2, L3 =1( negative logic ); no sensor data detected= tank is empty and the process will be stopped.
7. If the L1, L2=1; L3=0, then the pump speed in the primary will be reduced to 20%.
8. If the L1=1; L2=0; L3=0, then the pump speed is increased by 50%.
9. If the L1=0; L2=0; L3=0, then the pump speed is increased by 100%.
10. If the Nozzle position is 20<=x<=200, then it indicates a nozzle error and a message will be sent to the control unit for manual verification as shown in Fig 3
11. The entire process is instantaneously controlled by the fusion of sensor data.

VII. RESULT

1. A smart liquid filling machine that adjusts the motor speed based on time.it can also auto-adjust the motor speed in case of any change in the liquid level.
2. Auto-adjust system works on RF(random forest). this model is capable of predicting the water level based on the time(run time of system in minutes) and the motor speed
3. The dataset used to train this model consists of the run time of the system in minutes, the motor speed in %, and the water level.
4. The data can be represented in a scatter plot as shown below

Predicting the speed of the motor takes place with three different cases explained below as following

A. Case 1:Interval of 30 min

In case 1 the motor speed and liquid levels are measured at an interval of 30 min as shown in the image below. the results are shown in the below image. we got are accuracy is 0.986 And the mean error is 0.009234 which is the highest accuracy among the other 2 cases.

Conclusion

Overall its observed that the proposed solution has an overall higher accuracy of 98.6% and above under various testing parameter. This is achieved while maintaining a lower mean absolute error, as low as 0.009234. Considering the precision of the trained machine learning model, its observed from the datasets that motor speed is maintain ensuring efficiency and avoid motor overheating. Comparing the ANN model in various cases also reveals a lower loss fluctuations per epoch as compared to other solutions.

References

[1] Water plant optimization control system based on machine learning, Dongsheng Wanga,*, Yan Wanga , Rui Zhoua , Yong Caob , Fuchun Jiangb , Xue Zhangb , Jinghua Lia, doi: 10.5004/dwt.2021.27056, 222 (2021) 168–181. [2] Design and Implementation of a Water Level Controller using Fuzzy Logic, Namrata Dey Ria Mandal , M Monica Subashini, ISSN : 0975-4024 Vol 5 No 3 Jun-Jul 2013 2277 [3] Automatic Bottle Filling Machine, . B. Kalidasan1, J. Ben Ajai Raja International Research Journal of Engineering and Technology (IRJET) , Volume: 05 Issue: 03 | Mar-2018, p-ISSN: 2395-0072, [4] Automatic Bottle Filling System Using PLC Based Controller, Md. Liton Ahmed1, Shantonu Kundu, Journal of Advancement in Mechanics Volume 4 , 2019. [5] Automatic Vessel Filling System Using Raspberry-pi, Pallavi Vivek Shete, Minal V. Raundale, International Engineering Research Journal (IERJ) Volume 2 Issue 3 Page 1449-1451, 2016, ISSN 2395-1621 [6] Industrial Monitoring and Control Using Internet of Things, International Journal for Modern Trends in Science and Technology ,2016 [7] Design and Construct Intelligent Tank“Water Level Sensor, Journal of AL-Qadisiyah for computer science and mathematics Vol.10 No.3 Year 2018ISSN (Print): 2074 – 0204 ISSN (Online): 2521 – 350 [8] Development of an Ultrasonic Sensor Based Water Level Indicator With Pump Switching Technique, June 2018Journal of Electrical Engineering [9] Machine learning techniques for liquid level estimation using FBG temperature sensor array, Contents lists available at Science Direct Optical Fiber Technology [10] DC Motor Speed Control Using Machine Learning Algorithm, International Journal of Engineering Research & Technology (IJERT),ISSN: 2278-0181,2018 [11] Design Of Water Level Controller using Fuzzy Logic System , National Institute of Technology Rourkela , 2019 [12] Performance Comparison of Industrial Controller and Fuzzy Logic Controller For Water Level Tank Control System, International Research Journal of Modernization in Engineering Technology and Science ( Peer-Reviewed, Open Access, Fully Refereed International Journal ),2022 [13] Industrial Monitoring and Control Applications Using IOT, Shrushti Panpaliya,International Journal of Emerging Technologies in Engineering Research (IJETER) Volume 6, Issue 1, January (2018) [14] Secured Speech Industrial Automation Based on Raspberry Pi and IoT ,Bhagirathi Bai Vittal Rao 1* Hanume Gowda Chandrakanth ,International Journal of Intelligent Engineering and Systems, Vol.11, No.6, 2018 DOI: 10.22266/ijies2018.1231.23 [15] IoT based Machine Learning Automation Algorithm for Controlling the Industrial Loads ,Bhagirathi Bai Vittal Rao1* Hanume Gowda Chandrakanth,International Journal of Intelligent Engineering and Systems, Vol.14, No.4, 2021 DOI: 10.22266/ijies2021.0831.14 .

Copyright

Copyright © 2022 Bhagirathi Bai V, Hanume Gowda Chandrakanth. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.