Decision Tree Algorithm For Control Of Compressor Multiset .

Decision tree algorithm for control of compressormultiset in refrigeration industryIvan Šulekić*, Damir Milinković ** and Tomislav Špoljarić**University of Applied Sciences – Department of Electrical Engineering, Zagreb, Croatia** Ekonerg d.o.o., Koranska ul. 5, Zagreb, Croatiaisulekic@tvz.hr, tspoljaric@tvz.hrAbstract - This paper describes novel control algorithmfor the multi-compressor set. Algorithm has inputparameters that are collected from each compressor in theset. According to these parameters, algorithm decides whichcombination of compressors is the best for optimumperformance of the entire setKeywords - multi-compressor set, controls, weight factorparameters, Siemens, maximum efficiencyI.INTRODUCTIONCompressor multiset comprises of four compressors.Each of them serve to maintain pressure in the refrigerationsystem. Multisets control system monitors current pressurein the system and vital compressor parameters. If pressurein the system drops below referent value algorithm checksif next compressor is ready. If it is ready, control systemturns it on. During exploitation, compressors parameterscan dangerously exceed limit values and then controlsystem needs to turn that compressors off. Fig. 1 showsprogram algorithm for control system.Control system assigns weight factors to all monitoredparameters. With weight factors programmer can changeinfluence of every parameter on the reliability of a system.Control system compares parameters for each compressorand decides the most optimal compressor for operation.II.PROCESS AND CONTROL SYSTEM DESCRIPTIONRefrigeration system needs to be supplied with constantpressure. A compressor multiset is used for that purpose.Before modernization, control system multiset worked inthe following way: only one compressor in the set wouldwork until it would overload. Then another compressorwould start.New control system uses novel control algorithm forselection of the most optimal compressor. The most optimalcompressor is the one with smallest total parameter value.Total parameter value is calculated from the individualparameter values of monitored compressor (1)𝑝𝑡𝑜𝑡 𝜆𝑝𝑜𝑡 𝜆𝑝𝑤𝑡 𝑝𝑛ℎ 4𝑥𝜆𝑝𝑚𝑐𝑏 ]4( )𝑝𝑡𝑜𝑡 𝑡𝑜𝑡𝑎𝑙 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑣𝑎𝑙𝑢𝑒𝑝𝑚𝑐𝑏 𝑚𝑜𝑡𝑜𝑟 𝑐𝑖𝑟𝑐𝑢𝑖𝑡 𝑏𝑟𝑎𝑘𝑒𝑟 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑣𝑎𝑙𝑢𝑒𝑝𝑤𝑡 𝑤𝑖𝑛𝑑𝑖𝑛𝑔 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑣𝑎𝑙𝑢𝑒𝑝𝑜𝑡 𝑜𝑖𝑙 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑣𝑎𝑙𝑢𝑒𝑝𝑛ℎ 𝑤𝑜𝑟𝑘𝑖𝑛𝑔 ℎ𝑜𝑢𝑟 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑣𝑎𝑙𝑢𝑒𝜆 𝑤𝑒𝑖𝑔ℎ𝑡 𝑓𝑎𝑘𝑡𝑜𝑟During operation, every parameter value can bemodified with weight factors. All parameter values in thealgorithm are normalized .in value range from 0 to 1. Motoroverload switch normalized value can only be two values:0 or 1. Oil and winding temperature and working hournormalized values are in the range from 0 to 1.Currentcontrol system is designed for four compressors in multiset,but it can be modified for any number of compressors.Fig. 1: control algorithmCurrent control system monitors these parameters:winding temperature, oil temperature. motor overloadswitch state and number of working hours.PT100 sensors are used for measuring winding and oiltemperature. Motor overload switch state is obtained fromNO and NC contacts. Control system software countsnumber of working hours. Digital outputs are used forstarting and stopping compressors.1314III.AUTOMATION CONTROL DESIGNDevices used in this system are: compressor multiset,temperature sensors, motor overload switch, Siemens PLC1200 industrial controller with analog and digital I/Omodule and HMI (Human machine interface). Compressormultiset consists of four compressors. Temperature sensorsare realized with: PT100 sensors. Motor overload switchMIPRO 2020/CIS

have NC and NO contacts for current state signal. SiemensPLC 1200 industrial controller is used as a systemregulator. Analog I/O module is used for interaction withtemperature and pressure sensors. Digital I/O module isused for detecting motor overload switch states and forstarting and stopping compressors. All devices aredocumented in technical documentation [3].Fig 3: Graphic representation of overall liabilityparameters. If some parameter has a greater impact oncompressors reliability it is possible to change its weight.With this we can obtain maximum efficiency in the system.IV.Fig 2: Graphic representation of all devices in the systemCompressor control operates according to winding andoil temperature, pump working hours and motor overloadswitch state. Winding and oil temperature are detected withtemperature sensors. Motor overload switch state isdetected with contact. Working hours are counted bycontrol system during compressor exploitation.If controlsystem detects higher temperature values than limit valuesfrom winding or oil that event will be logged in controlsystem. Control system also logs every motor overloadswitch failure and it counts compressors working hours. Allthose events are normalized as values in the range from 0to 1. The algorithm combines all those normalized valuesand calculates overall reliability of the compressor. If someother compressor has a better reliability, then controlsystem will switch to that compressor. Temperaturenormalized values are calculated according to (2):𝑝𝑜𝑡,𝑤𝑡 75 𝑡𝑐𝑢𝑟𝑟50Process is organized in cycles. Number of cyclesdepends on the number of compressors in multiset. Eachcycle has control algorithm for one compressor.For every cycle we have three main steps:1) Is pressure smaller or equal to refernce ?2) Check if next compressor is ready3) Turn on compressor1)( )Fig 4: Program section 1: Desired pressure𝑡𝑐𝑢𝑟𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 g hour normalized values is calculated accordingto the (3) and (4).𝑝𝑛ℎ ′ 𝑝𝑛ℎ𝑚𝑎𝑥 ���𝑟𝑝𝑛ℎ 1 𝑝𝑛ℎ ′PROGRAM SOLUTION IN LADDER LOGIC2)( )( )𝑝𝑛ℎ𝑚𝑎𝑥 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑤𝑜𝑟𝑘𝑖𝑛𝑔 ℎ𝑜𝑢𝑟𝑠𝑝𝑛ℎ𝑐𝑢𝑟𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑤𝑜𝑟𝑘𝑖𝑛𝑔 ℎ𝑜𝑢𝑟𝑠Motor overload switch normalized value can only be 0or 1. Overall reliability of compressor is calculated with (1):𝑝𝑡𝑜𝑡 𝜆𝑝𝑜𝑡 𝜆𝑝𝑤𝑡 𝜆𝑝𝑛ℎ 4𝑥𝜆𝑝𝑚𝑐𝑏4( )On Fig. 3 there is a graphic representation of overallreliability for all compressors in the set:Current status of all compressors parameters isvisualized with HMI. Operator can see the current state ofall compressors. Programmer can give priority to someFig 5: Program section 2: Compressors availabilityMIPRO 2020/CIS1315

3)Analog I/O moduleDeviceI/O signalNameAI1Winding temperature compressor 1AI2Winding temperature compressor 2AI3Winding temperature compressor 3AI4Winding temperature compressor 4AI5Oil temperature compressor 1AI6Oil temperature compressor 2AI7Oil temperature compressor 3AI8Oil temperature compressor 4The code programming is done in Siemens TIA portalwith Simatic Step 7 software package [1]. It is used forcommunication setup, definition of variables and codewriting.Fig 6: Program section 3: Start compressorBefore programming it is necessary to make a list of allused variables. The list includes physical and virtual I/Ovariables used later in the program and they are shown intable 1.TABLE 1digital I/O moduleDevice1316LIST OF VARIABLES IN LADDER PROGRAMI/O signalNameI1Local start/stopI2Local/remoteI3Compressor 1 readyI4Compressor 2 readyI5Compressor 3 readyI6Compressor 4 readyI7MOS activatedQ1Turn on compressor 1Q2Turn on compressor 2Q3Turn on compressor 3Q4Turn on compressor 4Q6Turn off compressor 1Q2Turn off compressor 2Q7Turn off compressor 3Q8Turn off compressor 4In first part of the code system needs to check pressure.If the pressure is smaller than desired, control system firstneeds to check how many compressors are ready foractivation. Winding temperature, oil temperature, numberof working hours and state of motor overload switch needto be in operating limits for compressor activation (see Fig7). When controller checks how many compressors areready it decides which compressor to activate. It selects,according to control algorithm, the most reliablecompressor. Most reliable compressor is the one that hasbiggest overall reliability. Example of overall reliabilitycalculation is given in Fig. 8. Calculation is based onformulas (1), (2), (3) and (4). After selection compressor isactivated. Controller waits for a specified time. Ifcompressor doesn’t reach desired pressure controller needsto start second most reliable compressor (see Fig 9). Thisprocess is repeated until desired pressure is reached. Everytime when compressor stops new reliability is calculatedfor it. Control system has the same program cycle for everycompressor.Fig 7: Example of program section 1: Compressors availabilityMIPRO 2020/CIS

TABLE 2 LIST OF VARIABLES IN USER INTERFACENameCompressor 1 readyCompressor 2 readyCompressor 3 readyCompressor 4 readyMOS activatedTurn on compressor 1Turn on compressor 2Fig 8: Example of program section 2: Overall reliability calculationTurn on compressor 3Turn on compressor 4Turn off compressor 1Turn off compressor 2Turn off compressor 3Turn off compressor 4Winding temperature compressor 1Winding temperature compressor 2Winding temperature compressor 3Winding temperature compressor 4Oil temperature compressor 1Oil temperature compressor 2Oil temperature compressor 3Fig 9: Example of program section 3: Turning on compressorOil temperature compressor 4Working hours compressor 1Working hours compressor 2Working hours compressor 3Working hours compressor 4Number of available compressorsCurrent reliability of compressor 1Current reliability of compressor 2Current reliability of compressor 3Fig. 10: Example of monitoring mask for one compressorV.USER INTERFACEUser interface is created with Siemens TIA portal [2].With this interface, operator can monitor operation of theentire compressor multiset. Observed parameters arevisualized for every compressor. During exploitation,programmer can change weights of monitored parameters.Current values of monitoring parameters can be seen onscreen (Fig. 10). Table 2 lists all variables and parametersthat can be accessed through user interface. User interfaceis connected with Siemens PLC 1200 industrial controller.MIPRO 2020/CISCurrent reliability of compressor 4Operator can only manually start compressor with userinterface but only if compressor is ready. All othervariables are only for monitoring purposes.1317