This thesis presents an implementation of Industrial IoT machine-to-machine communication between server and client which monitored through the Manufacturing Message Specification (MMS) protocol over blockchain architecture. A 1D Convolutional Neural Network (CNN) is used for event classification of the raw time-series sensor data which installed in the MMS server. It is intended to give clients information about the condition of the environment to prevent danger. The weight and the model structure of the training process are saved into a hierarchical data format (HDF5) and YAML format respectively, which later be used for the real-time testing process. The proposed 1D CNN architecture with standard scaler pre-processing overcomes the previous architecture framework in terms of accuracy up to 99.60% for training, while 92.52% for testing process. Also, the processing time which only requires 312 us/step. In addition, a real-time event classification based on sensor data is constructed in the implementation. Meanwhile, real objects¡¯ values from the sensors are converted into MMS objects¡¯ values inside a virtual manufacturing device (VMD). Each object is declared with its own unique object name. Then MMS objects' values will be stored to the private blockchain network using a static difficulty number. In the private blockchain network, registered MMS clients can access every newly mined block and its MMS objects¡¯ values. The data is secured inside the private blockchain network as only registered participants can access it. The proposed framework overcomes the security issues that arise in the MMS communication protocol and the previous work combining MMS with MQTT, such as replay attack and single point of failure.