eh + maw @ Quebec quandary solved by Fourier transform By Michel Ruel and John Gerry Power spectral density analysis software hits the spot at a paper mill. Centrifugal cleaners Process plants are reaping, large benefits from the widely available analysis tool power spectral density analysis (PSDA). PSDA uses a mathe- ‘matical technique, the Fourier transform, to ana- Iyze process controler signals. A Fourier analysis can determine the amplitade, frequency, phase, and a hose of other functions for each sinusoidal ‘component of a given waveform. Specifically, mysterious and hard-to-trace process cycling problems can be located and brought t0 5 ‘ation more quickly and, thus, efficient. ‘Stumped at hundreds of dollars per minute ‘The production from 2 newsprint paper sachine in a Quebec, Canada, paper mill was ‘out of specification and had to be rejected. The lost revenue and expense was pushing a million dollars per day. A 16-minute cyde in the basis weight, humidity, and dry weight was causing unacceptable product variability. The eyele did ‘not seem to be present atthe wet end ofthe paper machine. A graphic of the process is shown in Figure 1 “The paper machine receives pulp in three forms: kraft pulp, thermomechanical pulp (IMP), and broke pulp.hugst 1998» Tritech ‘Terminology and definitions co. controller output PSDA power spectral donsity analysis Py process variable TMP) thermomechanical pulp hasig the weight of a wel ream (500 sheets) ‘of paper, reported in Canada as ‘grams per square ‘meter of paper try the dry weight of wel the baper from -the oven _ humidity the ratio of water weight in the paper to the total weight of the paper: raft pulp - pulp derived from ‘wood chips using hheat and ‘chemicals thermo: ‘pulp manufactured ‘mechanical by grinding pulp | stripped logs into chips and expos- ing them to high- pressure steam “The broke is obsained from rejected product that has been sent to a beater where i is mixed with water and reduced to pulp. All the paper machines in he mill are fed with the same kraft pulp and TMR. The broke pulp is different on ‘ach machine, and, infact, each machine’ broke pulp is derived from its own out-of spec paper. ‘Thirty-six hous into the problem, icemained unsolved. In the face of this serious situation, the production superintendene created a special task force comprising the most experienced engincers and technicians in che mill. Bven retired person- rel with experience on this parccular paper ‘machine were brought in. Two topnotch paper- smakers ffom competitive mills were also hired in an exhaustive artempc to ative at 2 quick solu- tion, Manufacturers ofthe wires, Fels, and dryers sent people to help the team. ‘Two weeks of traditional troubleshooting fensucd unsuccesfily: During that time the fol- owing steps were taken: Every suspected loop was pput in manual mode for one hous. Each pump, | roll, beating, and all mechanical parts were checked. The ser points for all the ertical loops ‘were changed (one ata time). Operation changes and human factors were considered. Each of the three types of pulp was removed one ata tnt for one hour ‘Ac this poing, the task force decided ic nsoded a brand-new perspective. Outside help with a novel, fresh tool was brought onboard ‘What's your sine? Data was collected from the process ata point before the paper machine usinga data acquisition package and trending software. This dita, dis played below in Figure 2, was taken from the machine chest. The process variable (PV) mea sured isthe pulp level in the machine chest. The graphic for the PV closed-loop time response is the top graph and the bottom graph isthe con- troller output (CO). The CO is the set point for the chee pulp flow loops. From the time response, ic was hard to reach any conclusions, It appeared to be noisy process response data only. Using this time daa, software was used to per- form a power spectral density calculation, A Fe 2. The clsei-4oop tine response forthe machine chest Jel conta Figure 3. The power spec dersity ana of te controler catpatded es i patra Inte + tap 198 @ Fg 4. The pulp consistency contol data and the PSDA interpretation ‘ power spectrum displays the frequency content of a signal. If che signa i a sine wave, only one frequebcy is present. If the signal is whice noise, all frequencies are present. ‘The later condition is the desied state for 2 succesful and continuous process with no ‘cling, Using PSDA, each frequency component is found, and any dominant frequencies are high- lighted, In che spectrum pictured in Figure 3, the horizohral axis is the period or inverse frequency (frequency). ‘Voila! Crest le probleme In Figure 3, notice the large spike ar 1,000 seconds (16.5 minutes). This corresponded ‘exactly to the cycle soen in the rest of the mill, ‘As a check, data was collected from the pulp consistency loop at the output of the machint ches. The results are shown in Figure 4, ‘The left graph shows the closed-loop time response of the consistency (PV) and the output ofthe consistency controler (CO). Again, iis ifcue to reach any ‘condlusions from this time response data. Running 2 power special density calculation on the CO, | howevts, again shows a large spike at 1,000 seconds (ee Figure 4, right). Thise wo power spectral density graphs clearly show & problem eyeling ae 16 minutes. From this, the engineers concluded that the cycling problem was located before the paper machite. Homing in on the culprit Since che mill had three other paper machines that djd not have 16-minute eye problems, it ‘was decided to check the broke pulp circuit by temporarily shutting it of During the iniial troubleshooting, the broke citcuic as shut off, but only for 60 minutes. Ie had noe been long enough. This sort of human cor i often seen in such situations. Without advanjed tools ike PSDA. variability sources are difficult 9 spor over such I er. After wo hours, the process cycling had dis- appeared. The source of the problem was found! Produce was now within acceptable limits, and the only mystery remaining was the specific defaule within the broke circuie. Guilty as charged With the quandary now narrowly defined, clues from bygone days and experiences fell quickly into place. One member ofthe task force recalled that the problem roughly coincided with a faulty mixer in the broke chest. To temporarily compensate for the lass of a mixes, an old circu lating pump ‘vas used to move the stock around in the tank with pumping action, ‘The pump had been designed co pump low- consistency stock. But the recycled stock was high consistency. As the pump circulated the pap, the consistency became thickein the area of the pump intake, and the flow was gradually reduced to almost nothing. Ac the same time and at the opposite end of the chest, the pulp Was becoming diluted and moze able to flo. As the thick pulp mixed with the thin, the circulating pump gradually was able to pump more. The overall effect was a 16-minute pumping and nor- pumping cycle. To confirm this scenario, the circulating ‘pump was stopped, and the broke pulp was agzin initoduced ac the paper machine, No eyling was present in the system. Later, the pump was restarted. The cycling reappeared, ‘Technique has other applications Process loops have a significant impact on final product variability not only in paper mills but in other industries ‘Advanced tools like PSDA can quickly iden- tify the source of eyce problems. The techniques are powerful and easy to implement, By knowing the eye’ period, problems relating to vibration, loop tuning, loop interaction, and the mechanics are readily solved. 7 For online information ‘elating to paper produc. and the pulp and paper industry, conncet to these wo Web sites seriall gc-calenglishl industry/fabricheml and sw kappiorg/- Behind the byline Michel Ruel i registered pro- fesional electrical engineer (PE) He founded TOP Consrl Inc. (Canada) in 1995. He is a senior member of ISA as well a4 Hecan be ‘contacted at mruci@ropconite. ‘cam, John Gerry has a master’s degree in chemical engincering and is also a PE. He founded ExpesTune, Inc. in 1986 and is the president. He can be con- twcted via ExperTiane's Web site at wuvwexpertune.com, member of IEE