The permanent magnet (PM) of synchronous motors is susceptible to factors such as high temperature, reverse magnetic field, and impact vibration, resulting in demagnetization faults that directly affect the operational performance of machine. To address the prediction problem of time-dependent demagnetization, an improved subdomain analytical model is proposed and reapplied to predict magnetic fields considering demagnetization effects. The degree and distribution of demagnetization are innovatively described and defined using orthogonal design and statistical methods. Taking the 20/24 PM in-wheel motor (IWM) as an object, the variation of magnetic field and back electromotive force (EMF) under different demagnetization cases and degrees are compared and analyzed according to the Maxwell's theory and fieldcircuit coupling time-stepping finite element (FE) method. And the evolution rules and reconstruction characteristics of unbalanced electromagnetic force (UEF), cogging torque, and output torque are explored. The results demonstrate that the improved model predictions agree well with that of simulation results, and the amplitude of flux density and back EMF in demagnetized region are significantly lower compared to healthy conditions. Under the uniform demagnetization, the waveform of UEF, cogging torque, and output torque generated by the residual magnetic field is almost consistent with the healthy condition, with relatively lower peaks. In other cases, with the expansion of demagnetization degrees and demagnetized PM's numbers, the resulting UEF increases in a stepped pattern, and the cogging torque and output torque both increase to varying degrees.