Increased milking frequency (IMF) in early lactation is a time and cost-effective farm management practice to enhance profit in the dairy industry. The process involves milking cows more often in early lactation alone. On farms that milk twice daily (2X), early lactation cows are milked four times daily (4X) for 21 d postpartum. Cows produce significantly more milk during this timeframe and continue to have increased milk yield (MY) when returned to 2X milking for the remainder of lactation. The objective of this dissertation was to discover the physiological processes of early lactation IMF that cause increased MY throughout lactation. All studies involved unilateral frequent milking (UFM) with 2X and 4X udder halves for 21 d in early lactation. The first study manipulated milking interval (MI), or the time between milkings, during early lactation IMF. Cows were either milked on an even MI every 6 h or unevenly on a 9:3:9:3 h MI. Unevenly milked cows produced more milk on the final day of 4X treatment, but there was no significant difference in the increased MY carry-over effect between MI groups. Therefore, farmers can utilize any MI that fits their schedule and still achieve significantly enhanced profits. The second study aimed to infer the metabolic mechanisms of early lactation IMF that increase MY by comparing it to bovine somatotropin (bST). Cows that underwent early lactation IMF received bST at 80 DIM. Both IMF and bST treatments significantly enhanced MY, but there was no interaction or synergistic effect between treatments. We surmised that IMF and bST cause increased MY through different metabolic mechanisms since IMF functions locally and bST operates systemically. The final study analyzed mammary tissue from 2X and 4X udder halves on the final day of UFM treatment. The mechanism in which IMF enhanced MY involved increased protein levels of signal transducer and activator of transcription (STAT5), activated and total protein kinase B (Akt), and total extracellular signal-regulated kinase 1 and 2 (ERK1/2) and reduced protein levels of total mammalian target of rapamycin (mTOR) and total mitogen activated protein kinase (MAPK) in 4X udder halves compared to 2X. / Doctor of Philosophy / Increased milking frequency in early lactation is the process of milking cows more often for the first 3 weeks after calving. Cows not only produce more milk during this timeframe but continue to have elevated milk yield throughout lactation. This phenomenon is called the increased milk yield carry-over effect. This dissertation aimed to further enhance the increased milk yield carry-over effect of early lactation increased milking frequency. All studies utilized unilateral frequent milking with left udder halves milked twice daily and right udder halves milked four-times daily. The first study compared even and uneven milking intervals alongside early lactation increased milking frequency. The right udder halves of the even groups were milked every 6 hours. Cows in the uneven milking interval group were milked on a 9:3:9:3 hour interval. The uneven milking interval group produced more milk than even milking interval group on the final day of milking frequency treatment. However, there was no difference in milk yield between groups at any other time point. All cows had an increased milk yield carry-over effect throughout 300 days of lactation. In the second study, cows received bovine somatotropin in mid-lactation to observe possible synergistic effects in enhancing milk yield after early lactation increased milking frequency. Both the bovine somatotropin and increased milking frequency increased milk yield, but there was no synergistic effect when the two treatments were combined. The final study analyzed proteins within mammary tissue after 3 weeks of unilateral frequent milking. Udder halves milked four-times daily had significantly elevated activated and total STAT5, activated and total Akt, and total ERK1/2. Udder halves milked twice daily had elevated total MAPK and total mTOR. These findings helped to understand the metabolic functioning of increased milking frequency in early lactation that causes a persistent increase in milk yield throughout lactation.

Das Ziel der Arbeit bestand darin zu untersuchen, inwieweit durch Kombination von einfach zu erhebenden Daten des Melkprozesses mittels Fuzzy Logic die Bewertung der Eutergesundheit in Automatischen Melksystemen (AMS) verbessert werden kann. Eigene Forschungen richteten sich auf die Evaluierung unterschiedlicher Milch- bzw. Melkparameter hinsichtlich ihrer Eignung als Inputvariable verfeinerter Erkennungsmodelle für Euterkrankheiten am Melkroboter sowie auf die Entwicklung und Optimierung von Fuzzy Logic Modellen. Im Rahmen eines Vorversuches ist geprüft worden, welche Variablen Eingang in das Erkennungsmodell finden sollten. Zu diesem Zweck wurden neben der Erhebung von Prozessdaten eines AMS Astronaut(r) am 26. und 27. Juli 2001 mit Geräten des Typs LactoCorder(r) low flow insgesamt 754 Einzelmessungen an Eutervierteln durchgeführt und verarbeitet. Die Auswertung dieses Datenmaterials ergab, dass zwischen "unauffälligen", d.h. mehr oder weniger gesunden, und "auffälligen", d.h. mehr oder minder kranken Eutervierteln gewisse Unterschiede hinsichtlich der viertelbezogenen Parameter normierte Milchbildungsrate und normierter Milchfluss sowie der normierten Zwischenmelkzeit bestehen, welche für eine Modellierung nutzbar sind. Auch die elektrische Leitfähigkeit sollte in das weitere Vorgehen einbezogen werden. Die Funktionalität einfacher Fuzzy Logic Modelle ist anhand von 74 erstellten Datensätzen erfolgreich getestet worden. Vom 4. bis 9. März 2002 fand in einem Betrieb mit zwei parallel arbeitenden AMS Astronaut(r) in einer Herde von 103 melkenden Kühen die Hauptuntersuchung statt. 4 282 Messungen mit LactoCorder(r) low flow-Geräten an 359 Eutervierteln sind einzeln analysiert worden. Nach Entfernung eventuell fehlerhafter oder nicht eindeutig zuzuordnender Datensätze verblieb eine nutzbare Datenbasis von 2 826 Datensätzen, welche an 195 "unauffälligen" und an 41 "auffälligen" Eutervierteln gemessen worden waren. Die statistische Aufbereitung der absoluten Parameterwerte ergab hoch signifikante Differenzen zwischen "unauffälligen" und "auffälligen" Vierteln für die untersuchten Eingangsvariablen Milchbildungsrate, Milchfluss, Zwischenmelkzeit und Leitfähigkeit. Zur weiteren Modellierung wurden 527 aus den Rohdaten berechnete, normierte Datensätze verwendet, weil diese weitgehend unbeeinflusst von störenden individuellen und Umwelteinflüssen sind. Es sind drei verschiedene Ansätze zur Modellierung der normierten Parameter verfolgt, eingehend analysiert und untereinander sowie mit den Alarmmeldungen der AMS verglichen worden. Während Ein-Parameter-Modelle die höchsten Spezifitätswerte aufzeigten, konnten für die Verknüpfung der vier Eingangsvariablen durch Indexmodelle die besten Sensitivitätswerte ermittelt werden. Das optimale Ergebnis erzielte jedoch die Modellierung mit Hilfe von Fuzzy Logic. Sie ergab mit 5,9 Prozent die niedrigste statistische Wahrscheinlichkeit von Fehlklassifizierungen (falschen Diagnosen). Die Alarmliste der AMS wies zwar eine ähnlich geringe Fehlerquote auf, war jedoch zugleich durch eine schlechte Sensitivität charakterisiert. In einer abschließenden vergleichenden Diskussion mit aktuellen Forschungsergebnissen bestätigte sich die Richtigkeit beider Grundannahmen der vorgelegten Dissertation: 1. Aus Daten des Melkprozesses zu ermittelnde, viertelbezogene Parameter eignen sich zur Ergänzung von Leitfähigkeitsmessungen für die Erkennung von Euterkrankheiten. 2. Fuzzy Logic ist eine handliche und wirkungsvolle Methodik zur Modellierung der Zusammenhänge zwischen Melk- und Milchparametern einerseits und der Eutergesundheit andererseits. / The goal of the work was to examine to what extent the evaluation of the udder health in Automatic Milking Systems (AMS) can be improved by the combination of simply measurable data of the milking process by means of Fuzzy Logic. Own investigations were centred on the evaluation of different milking and milk parameters regarding their suitability as input variable of recognition models for udder diseases as well as on the development and optimisation of Fuzzy Logic models. In the context of a preliminary study, it was examined which variables should form the input of the recognition model. For this purpose, besides the collection of the process data of an AMS Astronaut(r), 754 measurements of single udder quarters were accomplished with measuring devices LactoCorder(r) low flow and were individually processed. The evaluation of this data resulted in certain differences between "inconspicuous" (healthy) and "conspicuous" (more ore less ill) udder quarters regarding the quarter-related parameters milk production rate, milk flow and milking interval which are usable for model building. Also the electrical conductivity should be included into further procedures. The functionality of simple Fuzzy logic models was tested successfully on the basis of 74 generated data sets. The main investigation took place in a dairy farm with two parallel working AMS Astronaut(r) in a herd of 103 milking cows. 4,282 measurements at 359 udder quarters with devices LactoCorder(r) low flow were individually analysed. After removal of possibly incorrect data records or those which could not clearly be assigned remained a usable database of 2,826 records measured at 195 "inconspicuous" and at 41 "conspicuous" udder quarters. The statistic analysis of the absolute parameter values resulted in highly significant differences between "inconspicuous" and "conspicuous" quarters for the examined input variables milk production rate, milk flow, milking interval and electrical conductivity. For the further modelling, 527 standardised data sets computed from the raw data were used because these are largely uninfluenced by disturbing individual and environmental determinants. Three different approaches of model building for the standardised parameters were pursued, analysed and compared among themselves as well as with the alerts of the AMS. While the One-parameter-models pointed out the highest specificity values, the best sensitivity values were determined for the linkage of the four input variables by index-models. However, the optimal result of modelling was obtained by Fuzzy logic. It resulted in the lowest probability of false classifications (wrong diagnoses) with 5,9 per cent. The alarm list of the AMS featured a margin of error which was similar low, however, at the same time it was characterised by bad sensitivity. A concluding discussion with current investigation results confirmed the correctness of both basic assumptions of the dissertation: 1. Quarter-related parameters which can be calculated from data of the milking process are suitable as a complement to conductivity measurements for the detection of udder diseases. 2. Fuzzy Logic represents a manageable and effective methodology for modelling the interrelations between milking and milk parameters on the one hand and the udder health on the other hand.