Process control is of great importance in various sectors of textile industry. Yarn tension in rotor spinning is considered as a critical parameter and a few methods have been developed for on line measurement of this key parameter. Tension peaks have a detrimental effect on the quality of yarn and spinning stability. So, it is necessary to assess the interactions between process variables and tension peaks. Load cells as micro electro mechanical devices are widely used as process control tools. This work has adopted signal processing and tension load cells as measurement system. There is no reported work about using signal processing and strain gauge technology in determining the tension that are build up on the yarn during rotor spun yarn formation and this work is considered as a novel approach. An experimental set up is devised by using transducer, relevant circuits, amplifier and an interface to convert analog to digital. After system calibration, various tests were carried out by changing thick places, thin places, trash content and foreign matter content in the sliver as raw material variables. Also, yarn count and fibre type as product variables and rotor diameter, type of doffing nozzle and trash in rotor groove as machine variables were studied. The tension changes and tension peaks due to the interactions between the so called parameters were captured by this experimental setup. The results show that signal processing and strain gauge technology are powerful tools in measurement and control of yarn tension and tension peaks during rotor spinning. Higher yarn linear density, thick places, trash and foreign matter caused the increasement in the height of the tension peaks when the machine variables remained unchanged. That is while thin places, trash in rotor groove could decrease the tension peaks at the same conditions. Lower and higher spinning stability was the consequence of this increase and decrease on the height of tension peaks. Reduction in the rotor diameter when the rotational speed remained constant caused lower tensions on the yarn. This study revealed that the type of doffing nozzle and fibre type has not considerable effect on tensions that were build up on the yarn during spinning. Results were analyzed by ANOVA, Tukey statistical tests. A model was successfully developed by “decision tree” to predict and diagnose the defects on yarn. The results indicate that decision tree learned by J48 are powerful tools in prediction and diagnosing of the defects during yarn formation in rotor spinning respectively. Hence, the application of learned decision trees is suggested for the defect control in rotor spun yarn formation