Rolling　bearings　are　essential　for　machines　and　their　faults　can　cause　significant　losses.　Extracting　pulse　components　from　vibrations　is　difficult　due　to　complicated　working　conditions.　In　this　paper,　a　two-dimensional　overlapping　group　sparse　variation　method　based　on　non-convex　function　for　enhancing　time-frequency　modulation　bispectrum　characteristics　(OGSVMB)　to　identify　bearing　faults　is　proposed.　The　time-frequency　modulation　bispectrum　(TFMB)　demodulates　signals　into　bispectra,　highlighting　the　relationship　between　resonance　band　and　modulation　frequency.　A　mathematical　model　is　established　with　a　non-convex　penalty　term　to　constrain　the　distribution　of　two-dimensional　reconstructed　signal.　The　optimization-minimization　method　is　used　to　find　an　optimal　solution,　producing　a　bispectrum　with　clearer　modulation　characteristics　and　a　high　SNR　by　reducing　noise　and　irrelevant　components　while　preserving　the　group　sparsity　of　TFMB.　The　modulation　Gini　index　is　introduced　to　adaptively　determine　group　size.　Finally,　the　optimal　enhanced　envelope　spectrum　is　calculated　through　slice　optimization　to　identify　fault　characteristics.　The　effectiveness　of　the　proposed　method　in　diagnosing　rolling　bearing　faults　is　verified　through　simulation　and　experiments.　Compared　to　MSB,　original　TFMB,　Fast　Kurtogram,　and　Autogram,　the　OGSVMB　method　is　better　at　identifying　bearing　fault　characteristics.