In Industry 4.0, multiple research subjects have been involved in processing enormous information generated during the manufacturing process. Meanwhile, they all pursue the same goal, that is, to maintain the manufacturing process at good operating state. Hence, many researchers are focusing their researches on continuous health monitoring using digital signal processing techniques, in particular, most researches are focused on recognition of a fault condition only. However, by the time when a fault is recognised, at least one reject has been produced, or an impact to the system's operation has occurred. In addition, many algorithms are designed specifically for certain class of applications. This paper proposes a new system health condition indicator, Sum Standard Deviation of Frequency (SSDF), which is computed from a new computational process that segments raw data streams into time segments and the segments are synchrosqueezed continuous wavelet transformed. As long as sufficient data is available, SSDF shows distinct consecutive regions for "normal", "marginal" and "abnormal" machine conditions. Actions can then be taken while the machine is in "marginal" conditions in which the manufacturing quality is still acceptable. SSDF does not link to any application context information of the raw signal data stream hence making it context independent.